nfs41: reset the session slot table
[linux-2.6.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/utsname.h>
40 #include <linux/delay.h>
41 #include <linux/errno.h>
42 #include <linux/string.h>
43 #include <linux/sunrpc/clnt.h>
44 #include <linux/nfs.h>
45 #include <linux/nfs4.h>
46 #include <linux/nfs_fs.h>
47 #include <linux/nfs_page.h>
48 #include <linux/smp_lock.h>
49 #include <linux/namei.h>
50 #include <linux/mount.h>
51 #include <linux/module.h>
52
53 #include "nfs4_fs.h"
54 #include "delegation.h"
55 #include "internal.h"
56 #include "iostat.h"
57 #include "callback.h"
58
59 #define NFSDBG_FACILITY         NFSDBG_PROC
60
61 #define NFS4_POLL_RETRY_MIN     (HZ/10)
62 #define NFS4_POLL_RETRY_MAX     (15*HZ)
63
64 struct nfs4_opendata;
65 static int _nfs4_proc_open(struct nfs4_opendata *data);
66 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
67 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
68 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
69 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
70
71 /* Prevent leaks of NFSv4 errors into userland */
72 static int nfs4_map_errors(int err)
73 {
74         if (err < -1000) {
75                 dprintk("%s could not handle NFSv4 error %d\n",
76                                 __func__, -err);
77                 return -EIO;
78         }
79         return err;
80 }
81
82 /*
83  * This is our standard bitmap for GETATTR requests.
84  */
85 const u32 nfs4_fattr_bitmap[2] = {
86         FATTR4_WORD0_TYPE
87         | FATTR4_WORD0_CHANGE
88         | FATTR4_WORD0_SIZE
89         | FATTR4_WORD0_FSID
90         | FATTR4_WORD0_FILEID,
91         FATTR4_WORD1_MODE
92         | FATTR4_WORD1_NUMLINKS
93         | FATTR4_WORD1_OWNER
94         | FATTR4_WORD1_OWNER_GROUP
95         | FATTR4_WORD1_RAWDEV
96         | FATTR4_WORD1_SPACE_USED
97         | FATTR4_WORD1_TIME_ACCESS
98         | FATTR4_WORD1_TIME_METADATA
99         | FATTR4_WORD1_TIME_MODIFY
100 };
101
102 const u32 nfs4_statfs_bitmap[2] = {
103         FATTR4_WORD0_FILES_AVAIL
104         | FATTR4_WORD0_FILES_FREE
105         | FATTR4_WORD0_FILES_TOTAL,
106         FATTR4_WORD1_SPACE_AVAIL
107         | FATTR4_WORD1_SPACE_FREE
108         | FATTR4_WORD1_SPACE_TOTAL
109 };
110
111 const u32 nfs4_pathconf_bitmap[2] = {
112         FATTR4_WORD0_MAXLINK
113         | FATTR4_WORD0_MAXNAME,
114         0
115 };
116
117 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
118                         | FATTR4_WORD0_MAXREAD
119                         | FATTR4_WORD0_MAXWRITE
120                         | FATTR4_WORD0_LEASE_TIME,
121                         0
122 };
123
124 const u32 nfs4_fs_locations_bitmap[2] = {
125         FATTR4_WORD0_TYPE
126         | FATTR4_WORD0_CHANGE
127         | FATTR4_WORD0_SIZE
128         | FATTR4_WORD0_FSID
129         | FATTR4_WORD0_FILEID
130         | FATTR4_WORD0_FS_LOCATIONS,
131         FATTR4_WORD1_MODE
132         | FATTR4_WORD1_NUMLINKS
133         | FATTR4_WORD1_OWNER
134         | FATTR4_WORD1_OWNER_GROUP
135         | FATTR4_WORD1_RAWDEV
136         | FATTR4_WORD1_SPACE_USED
137         | FATTR4_WORD1_TIME_ACCESS
138         | FATTR4_WORD1_TIME_METADATA
139         | FATTR4_WORD1_TIME_MODIFY
140         | FATTR4_WORD1_MOUNTED_ON_FILEID
141 };
142
143 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
144                 struct nfs4_readdir_arg *readdir)
145 {
146         __be32 *start, *p;
147
148         BUG_ON(readdir->count < 80);
149         if (cookie > 2) {
150                 readdir->cookie = cookie;
151                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
152                 return;
153         }
154
155         readdir->cookie = 0;
156         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
157         if (cookie == 2)
158                 return;
159         
160         /*
161          * NFSv4 servers do not return entries for '.' and '..'
162          * Therefore, we fake these entries here.  We let '.'
163          * have cookie 0 and '..' have cookie 1.  Note that
164          * when talking to the server, we always send cookie 0
165          * instead of 1 or 2.
166          */
167         start = p = kmap_atomic(*readdir->pages, KM_USER0);
168         
169         if (cookie == 0) {
170                 *p++ = xdr_one;                                  /* next */
171                 *p++ = xdr_zero;                   /* cookie, first word */
172                 *p++ = xdr_one;                   /* cookie, second word */
173                 *p++ = xdr_one;                             /* entry len */
174                 memcpy(p, ".\0\0\0", 4);                        /* entry */
175                 p++;
176                 *p++ = xdr_one;                         /* bitmap length */
177                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
178                 *p++ = htonl(8);              /* attribute buffer length */
179                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
180         }
181         
182         *p++ = xdr_one;                                  /* next */
183         *p++ = xdr_zero;                   /* cookie, first word */
184         *p++ = xdr_two;                   /* cookie, second word */
185         *p++ = xdr_two;                             /* entry len */
186         memcpy(p, "..\0\0", 4);                         /* entry */
187         p++;
188         *p++ = xdr_one;                         /* bitmap length */
189         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
190         *p++ = htonl(8);              /* attribute buffer length */
191         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
192
193         readdir->pgbase = (char *)p - (char *)start;
194         readdir->count -= readdir->pgbase;
195         kunmap_atomic(start, KM_USER0);
196 }
197
198 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
199 {
200         int res;
201
202         might_sleep();
203
204         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
205                         nfs_wait_bit_killable, TASK_KILLABLE);
206         return res;
207 }
208
209 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
210 {
211         int res = 0;
212
213         might_sleep();
214
215         if (*timeout <= 0)
216                 *timeout = NFS4_POLL_RETRY_MIN;
217         if (*timeout > NFS4_POLL_RETRY_MAX)
218                 *timeout = NFS4_POLL_RETRY_MAX;
219         schedule_timeout_killable(*timeout);
220         if (fatal_signal_pending(current))
221                 res = -ERESTARTSYS;
222         *timeout <<= 1;
223         return res;
224 }
225
226 /* This is the error handling routine for processes that are allowed
227  * to sleep.
228  */
229 static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
230 {
231         struct nfs_client *clp = server->nfs_client;
232         struct nfs4_state *state = exception->state;
233         int ret = errorcode;
234
235         exception->retry = 0;
236         switch(errorcode) {
237                 case 0:
238                         return 0;
239                 case -NFS4ERR_ADMIN_REVOKED:
240                 case -NFS4ERR_BAD_STATEID:
241                 case -NFS4ERR_OPENMODE:
242                         if (state == NULL)
243                                 break;
244                         nfs4_state_mark_reclaim_nograce(clp, state);
245                 case -NFS4ERR_STALE_CLIENTID:
246                 case -NFS4ERR_STALE_STATEID:
247                 case -NFS4ERR_EXPIRED:
248                         nfs4_schedule_state_recovery(clp);
249                         ret = nfs4_wait_clnt_recover(clp);
250                         if (ret == 0)
251                                 exception->retry = 1;
252                         break;
253                 case -NFS4ERR_FILE_OPEN:
254                 case -NFS4ERR_GRACE:
255                 case -NFS4ERR_DELAY:
256                         ret = nfs4_delay(server->client, &exception->timeout);
257                         if (ret != 0)
258                                 break;
259                 case -NFS4ERR_OLD_STATEID:
260                         exception->retry = 1;
261         }
262         /* We failed to handle the error */
263         return nfs4_map_errors(ret);
264 }
265
266
267 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
268 {
269         struct nfs_client *clp = server->nfs_client;
270         spin_lock(&clp->cl_lock);
271         if (time_before(clp->cl_last_renewal,timestamp))
272                 clp->cl_last_renewal = timestamp;
273         spin_unlock(&clp->cl_lock);
274 }
275
276 #if defined(CONFIG_NFS_V4_1)
277
278 /*
279  * nfs4_free_slot - free a slot and efficiently update slot table.
280  *
281  * freeing a slot is trivially done by clearing its respective bit
282  * in the bitmap.
283  * If the freed slotid equals highest_used_slotid we want to update it
284  * so that the server would be able to size down the slot table if needed,
285  * otherwise we know that the highest_used_slotid is still in use.
286  * When updating highest_used_slotid there may be "holes" in the bitmap
287  * so we need to scan down from highest_used_slotid to 0 looking for the now
288  * highest slotid in use.
289  * If none found, highest_used_slotid is set to -1.
290  */
291 static void
292 nfs4_free_slot(struct nfs4_slot_table *tbl, u8 free_slotid)
293 {
294         int slotid = free_slotid;
295
296         spin_lock(&tbl->slot_tbl_lock);
297         /* clear used bit in bitmap */
298         __clear_bit(slotid, tbl->used_slots);
299
300         /* update highest_used_slotid when it is freed */
301         if (slotid == tbl->highest_used_slotid) {
302                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
303                 if (slotid >= 0 && slotid < tbl->max_slots)
304                         tbl->highest_used_slotid = slotid;
305                 else
306                         tbl->highest_used_slotid = -1;
307         }
308         rpc_wake_up_next(&tbl->slot_tbl_waitq);
309         spin_unlock(&tbl->slot_tbl_lock);
310         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
311                 free_slotid, tbl->highest_used_slotid);
312 }
313
314 void nfs41_sequence_free_slot(const struct nfs_client *clp,
315                               struct nfs4_sequence_res *res)
316 {
317         struct nfs4_slot_table *tbl;
318
319         if (!nfs4_has_session(clp)) {
320                 dprintk("%s: No session\n", __func__);
321                 return;
322         }
323         tbl = &clp->cl_session->fc_slot_table;
324         if (res->sr_slotid == NFS4_MAX_SLOT_TABLE) {
325                 dprintk("%s: No slot\n", __func__);
326                 /* just wake up the next guy waiting since
327                  * we may have not consumed a slot after all */
328                 rpc_wake_up_next(&tbl->slot_tbl_waitq);
329                 return;
330         }
331         nfs4_free_slot(tbl, res->sr_slotid);
332         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
333 }
334
335 static void nfs41_sequence_done(struct nfs_client *clp,
336                                 struct nfs4_sequence_res *res,
337                                 int rpc_status)
338 {
339         unsigned long timestamp;
340         struct nfs4_slot_table *tbl;
341         struct nfs4_slot *slot;
342
343         /*
344          * sr_status remains 1 if an RPC level error occurred. The server
345          * may or may not have processed the sequence operation..
346          * Proceed as if the server received and processed the sequence
347          * operation.
348          */
349         if (res->sr_status == 1)
350                 res->sr_status = NFS_OK;
351
352         /* -ERESTARTSYS can result in skipping nfs41_sequence_setup */
353         if (res->sr_slotid == NFS4_MAX_SLOT_TABLE)
354                 goto out;
355
356         tbl = &clp->cl_session->fc_slot_table;
357         slot = tbl->slots + res->sr_slotid;
358
359         if (res->sr_status == 0) {
360                 /* Update the slot's sequence and clientid lease timer */
361                 ++slot->seq_nr;
362                 timestamp = res->sr_renewal_time;
363                 spin_lock(&clp->cl_lock);
364                 if (time_before(clp->cl_last_renewal, timestamp))
365                         clp->cl_last_renewal = timestamp;
366                 spin_unlock(&clp->cl_lock);
367                 return;
368         }
369 out:
370         /* The session may be reset by one of the error handlers. */
371         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
372         nfs41_sequence_free_slot(clp, res);
373 }
374
375 /*
376  * nfs4_find_slot - efficiently look for a free slot
377  *
378  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
379  * If found, we mark the slot as used, update the highest_used_slotid,
380  * and respectively set up the sequence operation args.
381  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
382  *
383  * Note: must be called with under the slot_tbl_lock.
384  */
385 static u8
386 nfs4_find_slot(struct nfs4_slot_table *tbl, struct rpc_task *task)
387 {
388         int slotid;
389         u8 ret_id = NFS4_MAX_SLOT_TABLE;
390         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
391
392         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
393                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
394                 tbl->max_slots);
395         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
396         if (slotid >= tbl->max_slots)
397                 goto out;
398         __set_bit(slotid, tbl->used_slots);
399         if (slotid > tbl->highest_used_slotid)
400                 tbl->highest_used_slotid = slotid;
401         ret_id = slotid;
402 out:
403         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
404                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
405         return ret_id;
406 }
407
408 static int nfs41_setup_sequence(struct nfs4_session *session,
409                                 struct nfs4_sequence_args *args,
410                                 struct nfs4_sequence_res *res,
411                                 int cache_reply,
412                                 struct rpc_task *task)
413 {
414         struct nfs4_slot *slot;
415         struct nfs4_slot_table *tbl;
416         u8 slotid;
417
418         dprintk("--> %s\n", __func__);
419         /* slot already allocated? */
420         if (res->sr_slotid != NFS4_MAX_SLOT_TABLE)
421                 return 0;
422
423         memset(res, 0, sizeof(*res));
424         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
425         tbl = &session->fc_slot_table;
426
427         spin_lock(&tbl->slot_tbl_lock);
428         slotid = nfs4_find_slot(tbl, task);
429         if (slotid == NFS4_MAX_SLOT_TABLE) {
430                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
431                 spin_unlock(&tbl->slot_tbl_lock);
432                 dprintk("<-- %s: no free slots\n", __func__);
433                 return -EAGAIN;
434         }
435         spin_unlock(&tbl->slot_tbl_lock);
436
437         slot = tbl->slots + slotid;
438         args->sa_session = session;
439         args->sa_slotid = slotid;
440         args->sa_cache_this = cache_reply;
441
442         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
443
444         res->sr_session = session;
445         res->sr_slotid = slotid;
446         res->sr_renewal_time = jiffies;
447         /*
448          * sr_status is only set in decode_sequence, and so will remain
449          * set to 1 if an rpc level failure occurs.
450          */
451         res->sr_status = 1;
452         return 0;
453 }
454
455 int nfs4_setup_sequence(struct nfs_client *clp,
456                         struct nfs4_sequence_args *args,
457                         struct nfs4_sequence_res *res,
458                         int cache_reply,
459                         struct rpc_task *task)
460 {
461         int ret = 0;
462
463         dprintk("--> %s clp %p session %p sr_slotid %d\n",
464                 __func__, clp, clp->cl_session, res->sr_slotid);
465
466         if (!nfs4_has_session(clp))
467                 goto out;
468         ret = nfs41_setup_sequence(clp->cl_session, args, res, cache_reply,
469                                    task);
470         if (ret != -EAGAIN) {
471                 /* terminate rpc task */
472                 task->tk_status = ret;
473                 task->tk_action = NULL;
474         }
475 out:
476         dprintk("<-- %s status=%d\n", __func__, ret);
477         return ret;
478 }
479
480 struct nfs41_call_sync_data {
481         struct nfs_client *clp;
482         struct nfs4_sequence_args *seq_args;
483         struct nfs4_sequence_res *seq_res;
484         int cache_reply;
485 };
486
487 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
488 {
489         struct nfs41_call_sync_data *data = calldata;
490
491         dprintk("--> %s data->clp->cl_session %p\n", __func__,
492                 data->clp->cl_session);
493         if (nfs4_setup_sequence(data->clp, data->seq_args,
494                                 data->seq_res, data->cache_reply, task))
495                 return;
496         rpc_call_start(task);
497 }
498
499 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
500 {
501         struct nfs41_call_sync_data *data = calldata;
502
503         nfs41_sequence_done(data->clp, data->seq_res, task->tk_status);
504         nfs41_sequence_free_slot(data->clp, data->seq_res);
505 }
506
507 struct rpc_call_ops nfs41_call_sync_ops = {
508         .rpc_call_prepare = nfs41_call_sync_prepare,
509         .rpc_call_done = nfs41_call_sync_done,
510 };
511
512 static int nfs4_call_sync_sequence(struct nfs_client *clp,
513                                    struct rpc_clnt *clnt,
514                                    struct rpc_message *msg,
515                                    struct nfs4_sequence_args *args,
516                                    struct nfs4_sequence_res *res,
517                                    int cache_reply)
518 {
519         int ret;
520         struct rpc_task *task;
521         struct nfs41_call_sync_data data = {
522                 .clp = clp,
523                 .seq_args = args,
524                 .seq_res = res,
525                 .cache_reply = cache_reply,
526         };
527         struct rpc_task_setup task_setup = {
528                 .rpc_client = clnt,
529                 .rpc_message = msg,
530                 .callback_ops = &nfs41_call_sync_ops,
531                 .callback_data = &data
532         };
533
534         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
535         task = rpc_run_task(&task_setup);
536         if (IS_ERR(task))
537                 ret = PTR_ERR(task);
538         else {
539                 ret = task->tk_status;
540                 rpc_put_task(task);
541         }
542         return ret;
543 }
544
545 int _nfs4_call_sync_session(struct nfs_server *server,
546                             struct rpc_message *msg,
547                             struct nfs4_sequence_args *args,
548                             struct nfs4_sequence_res *res,
549                             int cache_reply)
550 {
551         return nfs4_call_sync_sequence(server->nfs_client, server->client,
552                                        msg, args, res, cache_reply);
553 }
554
555 #endif /* CONFIG_NFS_V4_1 */
556
557 int _nfs4_call_sync(struct nfs_server *server,
558                     struct rpc_message *msg,
559                     struct nfs4_sequence_args *args,
560                     struct nfs4_sequence_res *res,
561                     int cache_reply)
562 {
563         args->sa_session = res->sr_session = NULL;
564         return rpc_call_sync(server->client, msg, 0);
565 }
566
567 #define nfs4_call_sync(server, msg, args, res, cache_reply) \
568         (server)->nfs_client->cl_call_sync((server), (msg), &(args)->seq_args, \
569                         &(res)->seq_res, (cache_reply))
570
571 static void nfs4_sequence_done(const struct nfs_server *server,
572                                struct nfs4_sequence_res *res, int rpc_status)
573 {
574 #ifdef CONFIG_NFS_V4_1
575         if (nfs4_has_session(server->nfs_client))
576                 nfs41_sequence_done(server->nfs_client, res, rpc_status);
577 #endif /* CONFIG_NFS_V4_1 */
578 }
579
580 /* no restart, therefore free slot here */
581 static void nfs4_sequence_done_free_slot(const struct nfs_server *server,
582                                          struct nfs4_sequence_res *res,
583                                          int rpc_status)
584 {
585         nfs4_sequence_done(server, res, rpc_status);
586         nfs4_sequence_free_slot(server->nfs_client, res);
587 }
588
589 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
590 {
591         struct nfs_inode *nfsi = NFS_I(dir);
592
593         spin_lock(&dir->i_lock);
594         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
595         if (!cinfo->atomic || cinfo->before != nfsi->change_attr)
596                 nfs_force_lookup_revalidate(dir);
597         nfsi->change_attr = cinfo->after;
598         spin_unlock(&dir->i_lock);
599 }
600
601 struct nfs4_opendata {
602         struct kref kref;
603         struct nfs_openargs o_arg;
604         struct nfs_openres o_res;
605         struct nfs_open_confirmargs c_arg;
606         struct nfs_open_confirmres c_res;
607         struct nfs_fattr f_attr;
608         struct nfs_fattr dir_attr;
609         struct path path;
610         struct dentry *dir;
611         struct nfs4_state_owner *owner;
612         struct nfs4_state *state;
613         struct iattr attrs;
614         unsigned long timestamp;
615         unsigned int rpc_done : 1;
616         int rpc_status;
617         int cancelled;
618 };
619
620
621 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
622 {
623         p->o_res.f_attr = &p->f_attr;
624         p->o_res.dir_attr = &p->dir_attr;
625         p->o_res.seqid = p->o_arg.seqid;
626         p->c_res.seqid = p->c_arg.seqid;
627         p->o_res.server = p->o_arg.server;
628         nfs_fattr_init(&p->f_attr);
629         nfs_fattr_init(&p->dir_attr);
630 }
631
632 static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
633                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
634                 const struct iattr *attrs)
635 {
636         struct dentry *parent = dget_parent(path->dentry);
637         struct inode *dir = parent->d_inode;
638         struct nfs_server *server = NFS_SERVER(dir);
639         struct nfs4_opendata *p;
640
641         p = kzalloc(sizeof(*p), GFP_KERNEL);
642         if (p == NULL)
643                 goto err;
644         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
645         if (p->o_arg.seqid == NULL)
646                 goto err_free;
647         p->path.mnt = mntget(path->mnt);
648         p->path.dentry = dget(path->dentry);
649         p->dir = parent;
650         p->owner = sp;
651         atomic_inc(&sp->so_count);
652         p->o_arg.fh = NFS_FH(dir);
653         p->o_arg.open_flags = flags;
654         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
655         p->o_arg.clientid = server->nfs_client->cl_clientid;
656         p->o_arg.id = sp->so_owner_id.id;
657         p->o_arg.name = &p->path.dentry->d_name;
658         p->o_arg.server = server;
659         p->o_arg.bitmask = server->attr_bitmask;
660         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
661         p->o_res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
662         if (flags & O_EXCL) {
663                 u32 *s = (u32 *) p->o_arg.u.verifier.data;
664                 s[0] = jiffies;
665                 s[1] = current->pid;
666         } else if (flags & O_CREAT) {
667                 p->o_arg.u.attrs = &p->attrs;
668                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
669         }
670         p->c_arg.fh = &p->o_res.fh;
671         p->c_arg.stateid = &p->o_res.stateid;
672         p->c_arg.seqid = p->o_arg.seqid;
673         nfs4_init_opendata_res(p);
674         kref_init(&p->kref);
675         return p;
676 err_free:
677         kfree(p);
678 err:
679         dput(parent);
680         return NULL;
681 }
682
683 static void nfs4_opendata_free(struct kref *kref)
684 {
685         struct nfs4_opendata *p = container_of(kref,
686                         struct nfs4_opendata, kref);
687
688         nfs_free_seqid(p->o_arg.seqid);
689         if (p->state != NULL)
690                 nfs4_put_open_state(p->state);
691         nfs4_put_state_owner(p->owner);
692         dput(p->dir);
693         path_put(&p->path);
694         kfree(p);
695 }
696
697 static void nfs4_opendata_put(struct nfs4_opendata *p)
698 {
699         if (p != NULL)
700                 kref_put(&p->kref, nfs4_opendata_free);
701 }
702
703 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
704 {
705         int ret;
706
707         ret = rpc_wait_for_completion_task(task);
708         return ret;
709 }
710
711 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
712 {
713         int ret = 0;
714
715         if (open_mode & O_EXCL)
716                 goto out;
717         switch (mode & (FMODE_READ|FMODE_WRITE)) {
718                 case FMODE_READ:
719                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0;
720                         break;
721                 case FMODE_WRITE:
722                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0;
723                         break;
724                 case FMODE_READ|FMODE_WRITE:
725                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0;
726         }
727 out:
728         return ret;
729 }
730
731 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
732 {
733         if ((delegation->type & fmode) != fmode)
734                 return 0;
735         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
736                 return 0;
737         nfs_mark_delegation_referenced(delegation);
738         return 1;
739 }
740
741 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
742 {
743         switch (fmode) {
744                 case FMODE_WRITE:
745                         state->n_wronly++;
746                         break;
747                 case FMODE_READ:
748                         state->n_rdonly++;
749                         break;
750                 case FMODE_READ|FMODE_WRITE:
751                         state->n_rdwr++;
752         }
753         nfs4_state_set_mode_locked(state, state->state | fmode);
754 }
755
756 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
757 {
758         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
759                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
760         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
761         switch (fmode) {
762                 case FMODE_READ:
763                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
764                         break;
765                 case FMODE_WRITE:
766                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
767                         break;
768                 case FMODE_READ|FMODE_WRITE:
769                         set_bit(NFS_O_RDWR_STATE, &state->flags);
770         }
771 }
772
773 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
774 {
775         write_seqlock(&state->seqlock);
776         nfs_set_open_stateid_locked(state, stateid, fmode);
777         write_sequnlock(&state->seqlock);
778 }
779
780 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
781 {
782         /*
783          * Protect the call to nfs4_state_set_mode_locked and
784          * serialise the stateid update
785          */
786         write_seqlock(&state->seqlock);
787         if (deleg_stateid != NULL) {
788                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
789                 set_bit(NFS_DELEGATED_STATE, &state->flags);
790         }
791         if (open_stateid != NULL)
792                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
793         write_sequnlock(&state->seqlock);
794         spin_lock(&state->owner->so_lock);
795         update_open_stateflags(state, fmode);
796         spin_unlock(&state->owner->so_lock);
797 }
798
799 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
800 {
801         struct nfs_inode *nfsi = NFS_I(state->inode);
802         struct nfs_delegation *deleg_cur;
803         int ret = 0;
804
805         fmode &= (FMODE_READ|FMODE_WRITE);
806
807         rcu_read_lock();
808         deleg_cur = rcu_dereference(nfsi->delegation);
809         if (deleg_cur == NULL)
810                 goto no_delegation;
811
812         spin_lock(&deleg_cur->lock);
813         if (nfsi->delegation != deleg_cur ||
814             (deleg_cur->type & fmode) != fmode)
815                 goto no_delegation_unlock;
816
817         if (delegation == NULL)
818                 delegation = &deleg_cur->stateid;
819         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
820                 goto no_delegation_unlock;
821
822         nfs_mark_delegation_referenced(deleg_cur);
823         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
824         ret = 1;
825 no_delegation_unlock:
826         spin_unlock(&deleg_cur->lock);
827 no_delegation:
828         rcu_read_unlock();
829
830         if (!ret && open_stateid != NULL) {
831                 __update_open_stateid(state, open_stateid, NULL, fmode);
832                 ret = 1;
833         }
834
835         return ret;
836 }
837
838
839 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
840 {
841         struct nfs_delegation *delegation;
842
843         rcu_read_lock();
844         delegation = rcu_dereference(NFS_I(inode)->delegation);
845         if (delegation == NULL || (delegation->type & fmode) == fmode) {
846                 rcu_read_unlock();
847                 return;
848         }
849         rcu_read_unlock();
850         nfs_inode_return_delegation(inode);
851 }
852
853 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
854 {
855         struct nfs4_state *state = opendata->state;
856         struct nfs_inode *nfsi = NFS_I(state->inode);
857         struct nfs_delegation *delegation;
858         int open_mode = opendata->o_arg.open_flags & O_EXCL;
859         fmode_t fmode = opendata->o_arg.fmode;
860         nfs4_stateid stateid;
861         int ret = -EAGAIN;
862
863         for (;;) {
864                 if (can_open_cached(state, fmode, open_mode)) {
865                         spin_lock(&state->owner->so_lock);
866                         if (can_open_cached(state, fmode, open_mode)) {
867                                 update_open_stateflags(state, fmode);
868                                 spin_unlock(&state->owner->so_lock);
869                                 goto out_return_state;
870                         }
871                         spin_unlock(&state->owner->so_lock);
872                 }
873                 rcu_read_lock();
874                 delegation = rcu_dereference(nfsi->delegation);
875                 if (delegation == NULL ||
876                     !can_open_delegated(delegation, fmode)) {
877                         rcu_read_unlock();
878                         break;
879                 }
880                 /* Save the delegation */
881                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
882                 rcu_read_unlock();
883                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
884                 if (ret != 0)
885                         goto out;
886                 ret = -EAGAIN;
887
888                 /* Try to update the stateid using the delegation */
889                 if (update_open_stateid(state, NULL, &stateid, fmode))
890                         goto out_return_state;
891         }
892 out:
893         return ERR_PTR(ret);
894 out_return_state:
895         atomic_inc(&state->count);
896         return state;
897 }
898
899 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
900 {
901         struct inode *inode;
902         struct nfs4_state *state = NULL;
903         struct nfs_delegation *delegation;
904         int ret;
905
906         if (!data->rpc_done) {
907                 state = nfs4_try_open_cached(data);
908                 goto out;
909         }
910
911         ret = -EAGAIN;
912         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
913                 goto err;
914         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
915         ret = PTR_ERR(inode);
916         if (IS_ERR(inode))
917                 goto err;
918         ret = -ENOMEM;
919         state = nfs4_get_open_state(inode, data->owner);
920         if (state == NULL)
921                 goto err_put_inode;
922         if (data->o_res.delegation_type != 0) {
923                 int delegation_flags = 0;
924
925                 rcu_read_lock();
926                 delegation = rcu_dereference(NFS_I(inode)->delegation);
927                 if (delegation)
928                         delegation_flags = delegation->flags;
929                 rcu_read_unlock();
930                 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
931                         nfs_inode_set_delegation(state->inode,
932                                         data->owner->so_cred,
933                                         &data->o_res);
934                 else
935                         nfs_inode_reclaim_delegation(state->inode,
936                                         data->owner->so_cred,
937                                         &data->o_res);
938         }
939
940         update_open_stateid(state, &data->o_res.stateid, NULL,
941                         data->o_arg.fmode);
942         iput(inode);
943 out:
944         return state;
945 err_put_inode:
946         iput(inode);
947 err:
948         return ERR_PTR(ret);
949 }
950
951 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
952 {
953         struct nfs_inode *nfsi = NFS_I(state->inode);
954         struct nfs_open_context *ctx;
955
956         spin_lock(&state->inode->i_lock);
957         list_for_each_entry(ctx, &nfsi->open_files, list) {
958                 if (ctx->state != state)
959                         continue;
960                 get_nfs_open_context(ctx);
961                 spin_unlock(&state->inode->i_lock);
962                 return ctx;
963         }
964         spin_unlock(&state->inode->i_lock);
965         return ERR_PTR(-ENOENT);
966 }
967
968 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
969 {
970         struct nfs4_opendata *opendata;
971
972         opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, 0, NULL);
973         if (opendata == NULL)
974                 return ERR_PTR(-ENOMEM);
975         opendata->state = state;
976         atomic_inc(&state->count);
977         return opendata;
978 }
979
980 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
981 {
982         struct nfs4_state *newstate;
983         int ret;
984
985         opendata->o_arg.open_flags = 0;
986         opendata->o_arg.fmode = fmode;
987         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
988         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
989         nfs4_init_opendata_res(opendata);
990         ret = _nfs4_proc_open(opendata);
991         if (ret != 0)
992                 return ret; 
993         newstate = nfs4_opendata_to_nfs4_state(opendata);
994         if (IS_ERR(newstate))
995                 return PTR_ERR(newstate);
996         nfs4_close_state(&opendata->path, newstate, fmode);
997         *res = newstate;
998         return 0;
999 }
1000
1001 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1002 {
1003         struct nfs4_state *newstate;
1004         int ret;
1005
1006         /* memory barrier prior to reading state->n_* */
1007         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1008         smp_rmb();
1009         if (state->n_rdwr != 0) {
1010                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1011                 if (ret != 0)
1012                         return ret;
1013                 if (newstate != state)
1014                         return -ESTALE;
1015         }
1016         if (state->n_wronly != 0) {
1017                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1018                 if (ret != 0)
1019                         return ret;
1020                 if (newstate != state)
1021                         return -ESTALE;
1022         }
1023         if (state->n_rdonly != 0) {
1024                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1025                 if (ret != 0)
1026                         return ret;
1027                 if (newstate != state)
1028                         return -ESTALE;
1029         }
1030         /*
1031          * We may have performed cached opens for all three recoveries.
1032          * Check if we need to update the current stateid.
1033          */
1034         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1035             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1036                 write_seqlock(&state->seqlock);
1037                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1038                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1039                 write_sequnlock(&state->seqlock);
1040         }
1041         return 0;
1042 }
1043
1044 /*
1045  * OPEN_RECLAIM:
1046  *      reclaim state on the server after a reboot.
1047  */
1048 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1049 {
1050         struct nfs_delegation *delegation;
1051         struct nfs4_opendata *opendata;
1052         fmode_t delegation_type = 0;
1053         int status;
1054
1055         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1056         if (IS_ERR(opendata))
1057                 return PTR_ERR(opendata);
1058         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1059         opendata->o_arg.fh = NFS_FH(state->inode);
1060         rcu_read_lock();
1061         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1062         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1063                 delegation_type = delegation->type;
1064         rcu_read_unlock();
1065         opendata->o_arg.u.delegation_type = delegation_type;
1066         status = nfs4_open_recover(opendata, state);
1067         nfs4_opendata_put(opendata);
1068         return status;
1069 }
1070
1071 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1072 {
1073         struct nfs_server *server = NFS_SERVER(state->inode);
1074         struct nfs4_exception exception = { };
1075         int err;
1076         do {
1077                 err = _nfs4_do_open_reclaim(ctx, state);
1078                 if (err != -NFS4ERR_DELAY)
1079                         break;
1080                 nfs4_handle_exception(server, err, &exception);
1081         } while (exception.retry);
1082         return err;
1083 }
1084
1085 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1086 {
1087         struct nfs_open_context *ctx;
1088         int ret;
1089
1090         ctx = nfs4_state_find_open_context(state);
1091         if (IS_ERR(ctx))
1092                 return PTR_ERR(ctx);
1093         ret = nfs4_do_open_reclaim(ctx, state);
1094         put_nfs_open_context(ctx);
1095         return ret;
1096 }
1097
1098 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1099 {
1100         struct nfs4_opendata *opendata;
1101         int ret;
1102
1103         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1104         if (IS_ERR(opendata))
1105                 return PTR_ERR(opendata);
1106         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1107         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1108                         sizeof(opendata->o_arg.u.delegation.data));
1109         ret = nfs4_open_recover(opendata, state);
1110         nfs4_opendata_put(opendata);
1111         return ret;
1112 }
1113
1114 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1115 {
1116         struct nfs4_exception exception = { };
1117         struct nfs_server *server = NFS_SERVER(state->inode);
1118         int err;
1119         do {
1120                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1121                 switch (err) {
1122                         case 0:
1123                                 return err;
1124                         case -NFS4ERR_STALE_CLIENTID:
1125                         case -NFS4ERR_STALE_STATEID:
1126                         case -NFS4ERR_EXPIRED:
1127                                 /* Don't recall a delegation if it was lost */
1128                                 nfs4_schedule_state_recovery(server->nfs_client);
1129                                 return err;
1130                 }
1131                 err = nfs4_handle_exception(server, err, &exception);
1132         } while (exception.retry);
1133         return err;
1134 }
1135
1136 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1137 {
1138         struct nfs4_opendata *data = calldata;
1139
1140         data->rpc_status = task->tk_status;
1141         if (RPC_ASSASSINATED(task))
1142                 return;
1143         if (data->rpc_status == 0) {
1144                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1145                                 sizeof(data->o_res.stateid.data));
1146                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1147                 renew_lease(data->o_res.server, data->timestamp);
1148                 data->rpc_done = 1;
1149         }
1150 }
1151
1152 static void nfs4_open_confirm_release(void *calldata)
1153 {
1154         struct nfs4_opendata *data = calldata;
1155         struct nfs4_state *state = NULL;
1156
1157         /* If this request hasn't been cancelled, do nothing */
1158         if (data->cancelled == 0)
1159                 goto out_free;
1160         /* In case of error, no cleanup! */
1161         if (!data->rpc_done)
1162                 goto out_free;
1163         state = nfs4_opendata_to_nfs4_state(data);
1164         if (!IS_ERR(state))
1165                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1166 out_free:
1167         nfs4_opendata_put(data);
1168 }
1169
1170 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1171         .rpc_call_done = nfs4_open_confirm_done,
1172         .rpc_release = nfs4_open_confirm_release,
1173 };
1174
1175 /*
1176  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1177  */
1178 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1179 {
1180         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1181         struct rpc_task *task;
1182         struct  rpc_message msg = {
1183                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1184                 .rpc_argp = &data->c_arg,
1185                 .rpc_resp = &data->c_res,
1186                 .rpc_cred = data->owner->so_cred,
1187         };
1188         struct rpc_task_setup task_setup_data = {
1189                 .rpc_client = server->client,
1190                 .rpc_message = &msg,
1191                 .callback_ops = &nfs4_open_confirm_ops,
1192                 .callback_data = data,
1193                 .workqueue = nfsiod_workqueue,
1194                 .flags = RPC_TASK_ASYNC,
1195         };
1196         int status;
1197
1198         kref_get(&data->kref);
1199         data->rpc_done = 0;
1200         data->rpc_status = 0;
1201         data->timestamp = jiffies;
1202         task = rpc_run_task(&task_setup_data);
1203         if (IS_ERR(task))
1204                 return PTR_ERR(task);
1205         status = nfs4_wait_for_completion_rpc_task(task);
1206         if (status != 0) {
1207                 data->cancelled = 1;
1208                 smp_wmb();
1209         } else
1210                 status = data->rpc_status;
1211         rpc_put_task(task);
1212         return status;
1213 }
1214
1215 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1216 {
1217         struct nfs4_opendata *data = calldata;
1218         struct nfs4_state_owner *sp = data->owner;
1219
1220         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1221                 return;
1222         /*
1223          * Check if we still need to send an OPEN call, or if we can use
1224          * a delegation instead.
1225          */
1226         if (data->state != NULL) {
1227                 struct nfs_delegation *delegation;
1228
1229                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1230                         goto out_no_action;
1231                 rcu_read_lock();
1232                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1233                 if (delegation != NULL &&
1234                     test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) == 0) {
1235                         rcu_read_unlock();
1236                         goto out_no_action;
1237                 }
1238                 rcu_read_unlock();
1239         }
1240         /* Update sequence id. */
1241         data->o_arg.id = sp->so_owner_id.id;
1242         data->o_arg.clientid = sp->so_client->cl_clientid;
1243         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1244                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1245                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1246         }
1247         data->timestamp = jiffies;
1248         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
1249                                 &data->o_arg.seq_args,
1250                                 &data->o_res.seq_res, 1, task))
1251                 return;
1252         rpc_call_start(task);
1253         return;
1254 out_no_action:
1255         task->tk_action = NULL;
1256
1257 }
1258
1259 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1260 {
1261         struct nfs4_opendata *data = calldata;
1262
1263         data->rpc_status = task->tk_status;
1264
1265         nfs4_sequence_done_free_slot(data->o_arg.server, &data->o_res.seq_res,
1266                                      task->tk_status);
1267
1268         if (RPC_ASSASSINATED(task))
1269                 return;
1270         if (task->tk_status == 0) {
1271                 switch (data->o_res.f_attr->mode & S_IFMT) {
1272                         case S_IFREG:
1273                                 break;
1274                         case S_IFLNK:
1275                                 data->rpc_status = -ELOOP;
1276                                 break;
1277                         case S_IFDIR:
1278                                 data->rpc_status = -EISDIR;
1279                                 break;
1280                         default:
1281                                 data->rpc_status = -ENOTDIR;
1282                 }
1283                 renew_lease(data->o_res.server, data->timestamp);
1284                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1285                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1286         }
1287         data->rpc_done = 1;
1288 }
1289
1290 static void nfs4_open_release(void *calldata)
1291 {
1292         struct nfs4_opendata *data = calldata;
1293         struct nfs4_state *state = NULL;
1294
1295         /* If this request hasn't been cancelled, do nothing */
1296         if (data->cancelled == 0)
1297                 goto out_free;
1298         /* In case of error, no cleanup! */
1299         if (data->rpc_status != 0 || !data->rpc_done)
1300                 goto out_free;
1301         /* In case we need an open_confirm, no cleanup! */
1302         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1303                 goto out_free;
1304         state = nfs4_opendata_to_nfs4_state(data);
1305         if (!IS_ERR(state))
1306                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1307 out_free:
1308         nfs4_opendata_put(data);
1309 }
1310
1311 static const struct rpc_call_ops nfs4_open_ops = {
1312         .rpc_call_prepare = nfs4_open_prepare,
1313         .rpc_call_done = nfs4_open_done,
1314         .rpc_release = nfs4_open_release,
1315 };
1316
1317 /*
1318  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1319  */
1320 static int _nfs4_proc_open(struct nfs4_opendata *data)
1321 {
1322         struct inode *dir = data->dir->d_inode;
1323         struct nfs_server *server = NFS_SERVER(dir);
1324         struct nfs_openargs *o_arg = &data->o_arg;
1325         struct nfs_openres *o_res = &data->o_res;
1326         struct rpc_task *task;
1327         struct rpc_message msg = {
1328                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1329                 .rpc_argp = o_arg,
1330                 .rpc_resp = o_res,
1331                 .rpc_cred = data->owner->so_cred,
1332         };
1333         struct rpc_task_setup task_setup_data = {
1334                 .rpc_client = server->client,
1335                 .rpc_message = &msg,
1336                 .callback_ops = &nfs4_open_ops,
1337                 .callback_data = data,
1338                 .workqueue = nfsiod_workqueue,
1339                 .flags = RPC_TASK_ASYNC,
1340         };
1341         int status;
1342
1343         kref_get(&data->kref);
1344         data->rpc_done = 0;
1345         data->rpc_status = 0;
1346         data->cancelled = 0;
1347         task = rpc_run_task(&task_setup_data);
1348         if (IS_ERR(task))
1349                 return PTR_ERR(task);
1350         status = nfs4_wait_for_completion_rpc_task(task);
1351         if (status != 0) {
1352                 data->cancelled = 1;
1353                 smp_wmb();
1354         } else
1355                 status = data->rpc_status;
1356         rpc_put_task(task);
1357         if (status != 0 || !data->rpc_done)
1358                 return status;
1359
1360         if (o_res->fh.size == 0)
1361                 _nfs4_proc_lookup(dir, o_arg->name, &o_res->fh, o_res->f_attr);
1362
1363         if (o_arg->open_flags & O_CREAT) {
1364                 update_changeattr(dir, &o_res->cinfo);
1365                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1366         } else
1367                 nfs_refresh_inode(dir, o_res->dir_attr);
1368         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1369                 status = _nfs4_proc_open_confirm(data);
1370                 if (status != 0)
1371                         return status;
1372         }
1373         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1374                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1375         return 0;
1376 }
1377
1378 static int nfs4_recover_expired_lease(struct nfs_server *server)
1379 {
1380         struct nfs_client *clp = server->nfs_client;
1381         int ret;
1382
1383         for (;;) {
1384                 ret = nfs4_wait_clnt_recover(clp);
1385                 if (ret != 0)
1386                         return ret;
1387                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1388                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1389                         break;
1390                 nfs4_schedule_state_recovery(clp);
1391         }
1392         return 0;
1393 }
1394
1395 /*
1396  * OPEN_EXPIRED:
1397  *      reclaim state on the server after a network partition.
1398  *      Assumes caller holds the appropriate lock
1399  */
1400 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1401 {
1402         struct nfs4_opendata *opendata;
1403         int ret;
1404
1405         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1406         if (IS_ERR(opendata))
1407                 return PTR_ERR(opendata);
1408         ret = nfs4_open_recover(opendata, state);
1409         if (ret == -ESTALE)
1410                 d_drop(ctx->path.dentry);
1411         nfs4_opendata_put(opendata);
1412         return ret;
1413 }
1414
1415 static inline int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1416 {
1417         struct nfs_server *server = NFS_SERVER(state->inode);
1418         struct nfs4_exception exception = { };
1419         int err;
1420
1421         do {
1422                 err = _nfs4_open_expired(ctx, state);
1423                 if (err != -NFS4ERR_DELAY)
1424                         break;
1425                 nfs4_handle_exception(server, err, &exception);
1426         } while (exception.retry);
1427         return err;
1428 }
1429
1430 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1431 {
1432         struct nfs_open_context *ctx;
1433         int ret;
1434
1435         ctx = nfs4_state_find_open_context(state);
1436         if (IS_ERR(ctx))
1437                 return PTR_ERR(ctx);
1438         ret = nfs4_do_open_expired(ctx, state);
1439         put_nfs_open_context(ctx);
1440         return ret;
1441 }
1442
1443 /*
1444  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1445  * fields corresponding to attributes that were used to store the verifier.
1446  * Make sure we clobber those fields in the later setattr call
1447  */
1448 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1449 {
1450         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1451             !(sattr->ia_valid & ATTR_ATIME_SET))
1452                 sattr->ia_valid |= ATTR_ATIME;
1453
1454         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1455             !(sattr->ia_valid & ATTR_MTIME_SET))
1456                 sattr->ia_valid |= ATTR_MTIME;
1457 }
1458
1459 /*
1460  * Returns a referenced nfs4_state
1461  */
1462 static int _nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1463 {
1464         struct nfs4_state_owner  *sp;
1465         struct nfs4_state     *state = NULL;
1466         struct nfs_server       *server = NFS_SERVER(dir);
1467         struct nfs4_opendata *opendata;
1468         int status;
1469
1470         /* Protect against reboot recovery conflicts */
1471         status = -ENOMEM;
1472         if (!(sp = nfs4_get_state_owner(server, cred))) {
1473                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1474                 goto out_err;
1475         }
1476         status = nfs4_recover_expired_lease(server);
1477         if (status != 0)
1478                 goto err_put_state_owner;
1479         if (path->dentry->d_inode != NULL)
1480                 nfs4_return_incompatible_delegation(path->dentry->d_inode, fmode);
1481         status = -ENOMEM;
1482         opendata = nfs4_opendata_alloc(path, sp, fmode, flags, sattr);
1483         if (opendata == NULL)
1484                 goto err_put_state_owner;
1485
1486         if (path->dentry->d_inode != NULL)
1487                 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1488
1489         status = _nfs4_proc_open(opendata);
1490         if (status != 0)
1491                 goto err_opendata_put;
1492
1493         if (opendata->o_arg.open_flags & O_EXCL)
1494                 nfs4_exclusive_attrset(opendata, sattr);
1495
1496         state = nfs4_opendata_to_nfs4_state(opendata);
1497         status = PTR_ERR(state);
1498         if (IS_ERR(state))
1499                 goto err_opendata_put;
1500         nfs4_opendata_put(opendata);
1501         nfs4_put_state_owner(sp);
1502         *res = state;
1503         return 0;
1504 err_opendata_put:
1505         nfs4_opendata_put(opendata);
1506 err_put_state_owner:
1507         nfs4_put_state_owner(sp);
1508 out_err:
1509         *res = NULL;
1510         return status;
1511 }
1512
1513
1514 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1515 {
1516         struct nfs4_exception exception = { };
1517         struct nfs4_state *res;
1518         int status;
1519
1520         do {
1521                 status = _nfs4_do_open(dir, path, fmode, flags, sattr, cred, &res);
1522                 if (status == 0)
1523                         break;
1524                 /* NOTE: BAD_SEQID means the server and client disagree about the
1525                  * book-keeping w.r.t. state-changing operations
1526                  * (OPEN/CLOSE/LOCK/LOCKU...)
1527                  * It is actually a sign of a bug on the client or on the server.
1528                  *
1529                  * If we receive a BAD_SEQID error in the particular case of
1530                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1531                  * have unhashed the old state_owner for us, and that we can
1532                  * therefore safely retry using a new one. We should still warn
1533                  * the user though...
1534                  */
1535                 if (status == -NFS4ERR_BAD_SEQID) {
1536                         printk(KERN_WARNING "NFS: v4 server %s "
1537                                         " returned a bad sequence-id error!\n",
1538                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1539                         exception.retry = 1;
1540                         continue;
1541                 }
1542                 /*
1543                  * BAD_STATEID on OPEN means that the server cancelled our
1544                  * state before it received the OPEN_CONFIRM.
1545                  * Recover by retrying the request as per the discussion
1546                  * on Page 181 of RFC3530.
1547                  */
1548                 if (status == -NFS4ERR_BAD_STATEID) {
1549                         exception.retry = 1;
1550                         continue;
1551                 }
1552                 if (status == -EAGAIN) {
1553                         /* We must have found a delegation */
1554                         exception.retry = 1;
1555                         continue;
1556                 }
1557                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1558                                         status, &exception));
1559         } while (exception.retry);
1560         return res;
1561 }
1562
1563 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1564                             struct nfs_fattr *fattr, struct iattr *sattr,
1565                             struct nfs4_state *state)
1566 {
1567         struct nfs_server *server = NFS_SERVER(inode);
1568         struct nfs_setattrargs  arg = {
1569                 .fh             = NFS_FH(inode),
1570                 .iap            = sattr,
1571                 .server         = server,
1572                 .bitmask = server->attr_bitmask,
1573         };
1574         struct nfs_setattrres  res = {
1575                 .fattr          = fattr,
1576                 .server         = server,
1577         };
1578         struct rpc_message msg = {
1579                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1580                 .rpc_argp       = &arg,
1581                 .rpc_resp       = &res,
1582                 .rpc_cred       = cred,
1583         };
1584         unsigned long timestamp = jiffies;
1585         int status;
1586
1587         nfs_fattr_init(fattr);
1588
1589         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1590                 /* Use that stateid */
1591         } else if (state != NULL) {
1592                 nfs4_copy_stateid(&arg.stateid, state, current->files);
1593         } else
1594                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1595
1596         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
1597         if (status == 0 && state != NULL)
1598                 renew_lease(server, timestamp);
1599         return status;
1600 }
1601
1602 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1603                            struct nfs_fattr *fattr, struct iattr *sattr,
1604                            struct nfs4_state *state)
1605 {
1606         struct nfs_server *server = NFS_SERVER(inode);
1607         struct nfs4_exception exception = { };
1608         int err;
1609         do {
1610                 err = nfs4_handle_exception(server,
1611                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1612                                 &exception);
1613         } while (exception.retry);
1614         return err;
1615 }
1616
1617 struct nfs4_closedata {
1618         struct path path;
1619         struct inode *inode;
1620         struct nfs4_state *state;
1621         struct nfs_closeargs arg;
1622         struct nfs_closeres res;
1623         struct nfs_fattr fattr;
1624         unsigned long timestamp;
1625 };
1626
1627 static void nfs4_free_closedata(void *data)
1628 {
1629         struct nfs4_closedata *calldata = data;
1630         struct nfs4_state_owner *sp = calldata->state->owner;
1631
1632         nfs4_put_open_state(calldata->state);
1633         nfs_free_seqid(calldata->arg.seqid);
1634         nfs4_put_state_owner(sp);
1635         path_put(&calldata->path);
1636         kfree(calldata);
1637 }
1638
1639 static void nfs4_close_done(struct rpc_task *task, void *data)
1640 {
1641         struct nfs4_closedata *calldata = data;
1642         struct nfs4_state *state = calldata->state;
1643         struct nfs_server *server = NFS_SERVER(calldata->inode);
1644
1645         nfs4_sequence_done(server, &calldata->res.seq_res, task->tk_status);
1646         if (RPC_ASSASSINATED(task))
1647                 return;
1648         /* hmm. we are done with the inode, and in the process of freeing
1649          * the state_owner. we keep this around to process errors
1650          */
1651         switch (task->tk_status) {
1652                 case 0:
1653                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1654                         renew_lease(server, calldata->timestamp);
1655                         break;
1656                 case -NFS4ERR_STALE_STATEID:
1657                 case -NFS4ERR_OLD_STATEID:
1658                 case -NFS4ERR_BAD_STATEID:
1659                 case -NFS4ERR_EXPIRED:
1660                         if (calldata->arg.fmode == 0)
1661                                 break;
1662                 default:
1663                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN) {
1664                                 rpc_restart_call(task);
1665                                 return;
1666                         }
1667         }
1668         nfs4_sequence_free_slot(server->nfs_client, &calldata->res.seq_res);
1669         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1670 }
1671
1672 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1673 {
1674         struct nfs4_closedata *calldata = data;
1675         struct nfs4_state *state = calldata->state;
1676         int clear_rd, clear_wr, clear_rdwr;
1677
1678         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1679                 return;
1680
1681         clear_rd = clear_wr = clear_rdwr = 0;
1682         spin_lock(&state->owner->so_lock);
1683         /* Calculate the change in open mode */
1684         if (state->n_rdwr == 0) {
1685                 if (state->n_rdonly == 0) {
1686                         clear_rd |= test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1687                         clear_rdwr |= test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags);
1688                 }
1689                 if (state->n_wronly == 0) {
1690                         clear_wr |= test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1691                         clear_rdwr |= test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags);
1692                 }
1693         }
1694         spin_unlock(&state->owner->so_lock);
1695         if (!clear_rd && !clear_wr && !clear_rdwr) {
1696                 /* Note: exit _without_ calling nfs4_close_done */
1697                 task->tk_action = NULL;
1698                 return;
1699         }
1700         nfs_fattr_init(calldata->res.fattr);
1701         if (test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0) {
1702                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1703                 calldata->arg.fmode = FMODE_READ;
1704         } else if (test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0) {
1705                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1706                 calldata->arg.fmode = FMODE_WRITE;
1707         }
1708         calldata->timestamp = jiffies;
1709         if (nfs4_setup_sequence((NFS_SERVER(calldata->inode))->nfs_client,
1710                                 &calldata->arg.seq_args, &calldata->res.seq_res,
1711                                 1, task))
1712                 return;
1713         rpc_call_start(task);
1714 }
1715
1716 static const struct rpc_call_ops nfs4_close_ops = {
1717         .rpc_call_prepare = nfs4_close_prepare,
1718         .rpc_call_done = nfs4_close_done,
1719         .rpc_release = nfs4_free_closedata,
1720 };
1721
1722 /* 
1723  * It is possible for data to be read/written from a mem-mapped file 
1724  * after the sys_close call (which hits the vfs layer as a flush).
1725  * This means that we can't safely call nfsv4 close on a file until 
1726  * the inode is cleared. This in turn means that we are not good
1727  * NFSv4 citizens - we do not indicate to the server to update the file's 
1728  * share state even when we are done with one of the three share 
1729  * stateid's in the inode.
1730  *
1731  * NOTE: Caller must be holding the sp->so_owner semaphore!
1732  */
1733 int nfs4_do_close(struct path *path, struct nfs4_state *state, int wait)
1734 {
1735         struct nfs_server *server = NFS_SERVER(state->inode);
1736         struct nfs4_closedata *calldata;
1737         struct nfs4_state_owner *sp = state->owner;
1738         struct rpc_task *task;
1739         struct rpc_message msg = {
1740                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
1741                 .rpc_cred = state->owner->so_cred,
1742         };
1743         struct rpc_task_setup task_setup_data = {
1744                 .rpc_client = server->client,
1745                 .rpc_message = &msg,
1746                 .callback_ops = &nfs4_close_ops,
1747                 .workqueue = nfsiod_workqueue,
1748                 .flags = RPC_TASK_ASYNC,
1749         };
1750         int status = -ENOMEM;
1751
1752         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
1753         if (calldata == NULL)
1754                 goto out;
1755         calldata->inode = state->inode;
1756         calldata->state = state;
1757         calldata->arg.fh = NFS_FH(state->inode);
1758         calldata->arg.stateid = &state->open_stateid;
1759         /* Serialization for the sequence id */
1760         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid);
1761         if (calldata->arg.seqid == NULL)
1762                 goto out_free_calldata;
1763         calldata->arg.fmode = 0;
1764         calldata->arg.bitmask = server->cache_consistency_bitmask;
1765         calldata->res.fattr = &calldata->fattr;
1766         calldata->res.seqid = calldata->arg.seqid;
1767         calldata->res.server = server;
1768         calldata->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
1769         calldata->path.mnt = mntget(path->mnt);
1770         calldata->path.dentry = dget(path->dentry);
1771
1772         msg.rpc_argp = &calldata->arg,
1773         msg.rpc_resp = &calldata->res,
1774         task_setup_data.callback_data = calldata;
1775         task = rpc_run_task(&task_setup_data);
1776         if (IS_ERR(task))
1777                 return PTR_ERR(task);
1778         status = 0;
1779         if (wait)
1780                 status = rpc_wait_for_completion_task(task);
1781         rpc_put_task(task);
1782         return status;
1783 out_free_calldata:
1784         kfree(calldata);
1785 out:
1786         nfs4_put_open_state(state);
1787         nfs4_put_state_owner(sp);
1788         return status;
1789 }
1790
1791 static int nfs4_intent_set_file(struct nameidata *nd, struct path *path, struct nfs4_state *state, fmode_t fmode)
1792 {
1793         struct file *filp;
1794         int ret;
1795
1796         /* If the open_intent is for execute, we have an extra check to make */
1797         if (fmode & FMODE_EXEC) {
1798                 ret = nfs_may_open(state->inode,
1799                                 state->owner->so_cred,
1800                                 nd->intent.open.flags);
1801                 if (ret < 0)
1802                         goto out_close;
1803         }
1804         filp = lookup_instantiate_filp(nd, path->dentry, NULL);
1805         if (!IS_ERR(filp)) {
1806                 struct nfs_open_context *ctx;
1807                 ctx = nfs_file_open_context(filp);
1808                 ctx->state = state;
1809                 return 0;
1810         }
1811         ret = PTR_ERR(filp);
1812 out_close:
1813         nfs4_close_sync(path, state, fmode & (FMODE_READ|FMODE_WRITE));
1814         return ret;
1815 }
1816
1817 struct dentry *
1818 nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
1819 {
1820         struct path path = {
1821                 .mnt = nd->path.mnt,
1822                 .dentry = dentry,
1823         };
1824         struct dentry *parent;
1825         struct iattr attr;
1826         struct rpc_cred *cred;
1827         struct nfs4_state *state;
1828         struct dentry *res;
1829         fmode_t fmode = nd->intent.open.flags & (FMODE_READ | FMODE_WRITE | FMODE_EXEC);
1830
1831         if (nd->flags & LOOKUP_CREATE) {
1832                 attr.ia_mode = nd->intent.open.create_mode;
1833                 attr.ia_valid = ATTR_MODE;
1834                 if (!IS_POSIXACL(dir))
1835                         attr.ia_mode &= ~current_umask();
1836         } else {
1837                 attr.ia_valid = 0;
1838                 BUG_ON(nd->intent.open.flags & O_CREAT);
1839         }
1840
1841         cred = rpc_lookup_cred();
1842         if (IS_ERR(cred))
1843                 return (struct dentry *)cred;
1844         parent = dentry->d_parent;
1845         /* Protect against concurrent sillydeletes */
1846         nfs_block_sillyrename(parent);
1847         state = nfs4_do_open(dir, &path, fmode, nd->intent.open.flags, &attr, cred);
1848         put_rpccred(cred);
1849         if (IS_ERR(state)) {
1850                 if (PTR_ERR(state) == -ENOENT) {
1851                         d_add(dentry, NULL);
1852                         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
1853                 }
1854                 nfs_unblock_sillyrename(parent);
1855                 return (struct dentry *)state;
1856         }
1857         res = d_add_unique(dentry, igrab(state->inode));
1858         if (res != NULL)
1859                 path.dentry = res;
1860         nfs_set_verifier(path.dentry, nfs_save_change_attribute(dir));
1861         nfs_unblock_sillyrename(parent);
1862         nfs4_intent_set_file(nd, &path, state, fmode);
1863         return res;
1864 }
1865
1866 int
1867 nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, struct nameidata *nd)
1868 {
1869         struct path path = {
1870                 .mnt = nd->path.mnt,
1871                 .dentry = dentry,
1872         };
1873         struct rpc_cred *cred;
1874         struct nfs4_state *state;
1875         fmode_t fmode = openflags & (FMODE_READ | FMODE_WRITE);
1876
1877         cred = rpc_lookup_cred();
1878         if (IS_ERR(cred))
1879                 return PTR_ERR(cred);
1880         state = nfs4_do_open(dir, &path, fmode, openflags, NULL, cred);
1881         put_rpccred(cred);
1882         if (IS_ERR(state)) {
1883                 switch (PTR_ERR(state)) {
1884                         case -EPERM:
1885                         case -EACCES:
1886                         case -EDQUOT:
1887                         case -ENOSPC:
1888                         case -EROFS:
1889                                 lookup_instantiate_filp(nd, (struct dentry *)state, NULL);
1890                                 return 1;
1891                         default:
1892                                 goto out_drop;
1893                 }
1894         }
1895         if (state->inode == dentry->d_inode) {
1896                 nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
1897                 nfs4_intent_set_file(nd, &path, state, fmode);
1898                 return 1;
1899         }
1900         nfs4_close_sync(&path, state, fmode);
1901 out_drop:
1902         d_drop(dentry);
1903         return 0;
1904 }
1905
1906 void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
1907 {
1908         if (ctx->state == NULL)
1909                 return;
1910         if (is_sync)
1911                 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
1912         else
1913                 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
1914 }
1915
1916 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
1917 {
1918         struct nfs4_server_caps_arg args = {
1919                 .fhandle = fhandle,
1920         };
1921         struct nfs4_server_caps_res res = {};
1922         struct rpc_message msg = {
1923                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
1924                 .rpc_argp = &args,
1925                 .rpc_resp = &res,
1926         };
1927         int status;
1928
1929         status = nfs4_call_sync(server, &msg, &args, &res, 0);
1930         if (status == 0) {
1931                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
1932                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
1933                         server->caps |= NFS_CAP_ACLS;
1934                 if (res.has_links != 0)
1935                         server->caps |= NFS_CAP_HARDLINKS;
1936                 if (res.has_symlinks != 0)
1937                         server->caps |= NFS_CAP_SYMLINKS;
1938                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
1939                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
1940                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
1941                 server->acl_bitmask = res.acl_bitmask;
1942         }
1943
1944         return status;
1945 }
1946
1947 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
1948 {
1949         struct nfs4_exception exception = { };
1950         int err;
1951         do {
1952                 err = nfs4_handle_exception(server,
1953                                 _nfs4_server_capabilities(server, fhandle),
1954                                 &exception);
1955         } while (exception.retry);
1956         return err;
1957 }
1958
1959 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
1960                 struct nfs_fsinfo *info)
1961 {
1962         struct nfs4_lookup_root_arg args = {
1963                 .bitmask = nfs4_fattr_bitmap,
1964         };
1965         struct nfs4_lookup_res res = {
1966                 .server = server,
1967                 .fattr = info->fattr,
1968                 .fh = fhandle,
1969         };
1970         struct rpc_message msg = {
1971                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
1972                 .rpc_argp = &args,
1973                 .rpc_resp = &res,
1974         };
1975         nfs_fattr_init(info->fattr);
1976         return nfs4_call_sync(server, &msg, &args, &res, 0);
1977 }
1978
1979 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
1980                 struct nfs_fsinfo *info)
1981 {
1982         struct nfs4_exception exception = { };
1983         int err;
1984         do {
1985                 err = nfs4_handle_exception(server,
1986                                 _nfs4_lookup_root(server, fhandle, info),
1987                                 &exception);
1988         } while (exception.retry);
1989         return err;
1990 }
1991
1992 /*
1993  * get the file handle for the "/" directory on the server
1994  */
1995 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
1996                               struct nfs_fsinfo *info)
1997 {
1998         int status;
1999
2000         status = nfs4_lookup_root(server, fhandle, info);
2001         if (status == 0)
2002                 status = nfs4_server_capabilities(server, fhandle);
2003         if (status == 0)
2004                 status = nfs4_do_fsinfo(server, fhandle, info);
2005         return nfs4_map_errors(status);
2006 }
2007
2008 /*
2009  * Get locations and (maybe) other attributes of a referral.
2010  * Note that we'll actually follow the referral later when
2011  * we detect fsid mismatch in inode revalidation
2012  */
2013 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2014 {
2015         int status = -ENOMEM;
2016         struct page *page = NULL;
2017         struct nfs4_fs_locations *locations = NULL;
2018
2019         page = alloc_page(GFP_KERNEL);
2020         if (page == NULL)
2021                 goto out;
2022         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2023         if (locations == NULL)
2024                 goto out;
2025
2026         status = nfs4_proc_fs_locations(dir, name, locations, page);
2027         if (status != 0)
2028                 goto out;
2029         /* Make sure server returned a different fsid for the referral */
2030         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2031                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2032                 status = -EIO;
2033                 goto out;
2034         }
2035
2036         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2037         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2038         if (!fattr->mode)
2039                 fattr->mode = S_IFDIR;
2040         memset(fhandle, 0, sizeof(struct nfs_fh));
2041 out:
2042         if (page)
2043                 __free_page(page);
2044         if (locations)
2045                 kfree(locations);
2046         return status;
2047 }
2048
2049 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2050 {
2051         struct nfs4_getattr_arg args = {
2052                 .fh = fhandle,
2053                 .bitmask = server->attr_bitmask,
2054         };
2055         struct nfs4_getattr_res res = {
2056                 .fattr = fattr,
2057                 .server = server,
2058         };
2059         struct rpc_message msg = {
2060                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2061                 .rpc_argp = &args,
2062                 .rpc_resp = &res,
2063         };
2064         
2065         nfs_fattr_init(fattr);
2066         return nfs4_call_sync(server, &msg, &args, &res, 0);
2067 }
2068
2069 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2070 {
2071         struct nfs4_exception exception = { };
2072         int err;
2073         do {
2074                 err = nfs4_handle_exception(server,
2075                                 _nfs4_proc_getattr(server, fhandle, fattr),
2076                                 &exception);
2077         } while (exception.retry);
2078         return err;
2079 }
2080
2081 /* 
2082  * The file is not closed if it is opened due to the a request to change
2083  * the size of the file. The open call will not be needed once the
2084  * VFS layer lookup-intents are implemented.
2085  *
2086  * Close is called when the inode is destroyed.
2087  * If we haven't opened the file for O_WRONLY, we
2088  * need to in the size_change case to obtain a stateid.
2089  *
2090  * Got race?
2091  * Because OPEN is always done by name in nfsv4, it is
2092  * possible that we opened a different file by the same
2093  * name.  We can recognize this race condition, but we
2094  * can't do anything about it besides returning an error.
2095  *
2096  * This will be fixed with VFS changes (lookup-intent).
2097  */
2098 static int
2099 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2100                   struct iattr *sattr)
2101 {
2102         struct inode *inode = dentry->d_inode;
2103         struct rpc_cred *cred = NULL;
2104         struct nfs4_state *state = NULL;
2105         int status;
2106
2107         nfs_fattr_init(fattr);
2108         
2109         /* Search for an existing open(O_WRITE) file */
2110         if (sattr->ia_valid & ATTR_FILE) {
2111                 struct nfs_open_context *ctx;
2112
2113                 ctx = nfs_file_open_context(sattr->ia_file);
2114                 if (ctx) {
2115                         cred = ctx->cred;
2116                         state = ctx->state;
2117                 }
2118         }
2119
2120         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2121         if (status == 0)
2122                 nfs_setattr_update_inode(inode, sattr);
2123         return status;
2124 }
2125
2126 static int _nfs4_proc_lookupfh(struct nfs_server *server, const struct nfs_fh *dirfh,
2127                 const struct qstr *name, struct nfs_fh *fhandle,
2128                 struct nfs_fattr *fattr)
2129 {
2130         int                    status;
2131         struct nfs4_lookup_arg args = {
2132                 .bitmask = server->attr_bitmask,
2133                 .dir_fh = dirfh,
2134                 .name = name,
2135         };
2136         struct nfs4_lookup_res res = {
2137                 .server = server,
2138                 .fattr = fattr,
2139                 .fh = fhandle,
2140         };
2141         struct rpc_message msg = {
2142                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2143                 .rpc_argp = &args,
2144                 .rpc_resp = &res,
2145         };
2146
2147         nfs_fattr_init(fattr);
2148
2149         dprintk("NFS call  lookupfh %s\n", name->name);
2150         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2151         dprintk("NFS reply lookupfh: %d\n", status);
2152         return status;
2153 }
2154
2155 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2156                               struct qstr *name, struct nfs_fh *fhandle,
2157                               struct nfs_fattr *fattr)
2158 {
2159         struct nfs4_exception exception = { };
2160         int err;
2161         do {
2162                 err = _nfs4_proc_lookupfh(server, dirfh, name, fhandle, fattr);
2163                 /* FIXME: !!!! */
2164                 if (err == -NFS4ERR_MOVED) {
2165                         err = -EREMOTE;
2166                         break;
2167                 }
2168                 err = nfs4_handle_exception(server, err, &exception);
2169         } while (exception.retry);
2170         return err;
2171 }
2172
2173 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
2174                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2175 {
2176         int status;
2177         
2178         dprintk("NFS call  lookup %s\n", name->name);
2179         status = _nfs4_proc_lookupfh(NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2180         if (status == -NFS4ERR_MOVED)
2181                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2182         dprintk("NFS reply lookup: %d\n", status);
2183         return status;
2184 }
2185
2186 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2187 {
2188         struct nfs4_exception exception = { };
2189         int err;
2190         do {
2191                 err = nfs4_handle_exception(NFS_SERVER(dir),
2192                                 _nfs4_proc_lookup(dir, name, fhandle, fattr),
2193                                 &exception);
2194         } while (exception.retry);
2195         return err;
2196 }
2197
2198 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2199 {
2200         struct nfs_server *server = NFS_SERVER(inode);
2201         struct nfs_fattr fattr;
2202         struct nfs4_accessargs args = {
2203                 .fh = NFS_FH(inode),
2204                 .bitmask = server->attr_bitmask,
2205         };
2206         struct nfs4_accessres res = {
2207                 .server = server,
2208                 .fattr = &fattr,
2209         };
2210         struct rpc_message msg = {
2211                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2212                 .rpc_argp = &args,
2213                 .rpc_resp = &res,
2214                 .rpc_cred = entry->cred,
2215         };
2216         int mode = entry->mask;
2217         int status;
2218
2219         /*
2220          * Determine which access bits we want to ask for...
2221          */
2222         if (mode & MAY_READ)
2223                 args.access |= NFS4_ACCESS_READ;
2224         if (S_ISDIR(inode->i_mode)) {
2225                 if (mode & MAY_WRITE)
2226                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2227                 if (mode & MAY_EXEC)
2228                         args.access |= NFS4_ACCESS_LOOKUP;
2229         } else {
2230                 if (mode & MAY_WRITE)
2231                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2232                 if (mode & MAY_EXEC)
2233                         args.access |= NFS4_ACCESS_EXECUTE;
2234         }
2235         nfs_fattr_init(&fattr);
2236         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2237         if (!status) {
2238                 entry->mask = 0;
2239                 if (res.access & NFS4_ACCESS_READ)
2240                         entry->mask |= MAY_READ;
2241                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2242                         entry->mask |= MAY_WRITE;
2243                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2244                         entry->mask |= MAY_EXEC;
2245                 nfs_refresh_inode(inode, &fattr);
2246         }
2247         return status;
2248 }
2249
2250 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2251 {
2252         struct nfs4_exception exception = { };
2253         int err;
2254         do {
2255                 err = nfs4_handle_exception(NFS_SERVER(inode),
2256                                 _nfs4_proc_access(inode, entry),
2257                                 &exception);
2258         } while (exception.retry);
2259         return err;
2260 }
2261
2262 /*
2263  * TODO: For the time being, we don't try to get any attributes
2264  * along with any of the zero-copy operations READ, READDIR,
2265  * READLINK, WRITE.
2266  *
2267  * In the case of the first three, we want to put the GETATTR
2268  * after the read-type operation -- this is because it is hard
2269  * to predict the length of a GETATTR response in v4, and thus
2270  * align the READ data correctly.  This means that the GETATTR
2271  * may end up partially falling into the page cache, and we should
2272  * shift it into the 'tail' of the xdr_buf before processing.
2273  * To do this efficiently, we need to know the total length
2274  * of data received, which doesn't seem to be available outside
2275  * of the RPC layer.
2276  *
2277  * In the case of WRITE, we also want to put the GETATTR after
2278  * the operation -- in this case because we want to make sure
2279  * we get the post-operation mtime and size.  This means that
2280  * we can't use xdr_encode_pages() as written: we need a variant
2281  * of it which would leave room in the 'tail' iovec.
2282  *
2283  * Both of these changes to the XDR layer would in fact be quite
2284  * minor, but I decided to leave them for a subsequent patch.
2285  */
2286 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2287                 unsigned int pgbase, unsigned int pglen)
2288 {
2289         struct nfs4_readlink args = {
2290                 .fh       = NFS_FH(inode),
2291                 .pgbase   = pgbase,
2292                 .pglen    = pglen,
2293                 .pages    = &page,
2294         };
2295         struct nfs4_readlink_res res;
2296         struct rpc_message msg = {
2297                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2298                 .rpc_argp = &args,
2299                 .rpc_resp = &res,
2300         };
2301
2302         return nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
2303 }
2304
2305 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2306                 unsigned int pgbase, unsigned int pglen)
2307 {
2308         struct nfs4_exception exception = { };
2309         int err;
2310         do {
2311                 err = nfs4_handle_exception(NFS_SERVER(inode),
2312                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2313                                 &exception);
2314         } while (exception.retry);
2315         return err;
2316 }
2317
2318 /*
2319  * Got race?
2320  * We will need to arrange for the VFS layer to provide an atomic open.
2321  * Until then, this create/open method is prone to inefficiency and race
2322  * conditions due to the lookup, create, and open VFS calls from sys_open()
2323  * placed on the wire.
2324  *
2325  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2326  * The file will be opened again in the subsequent VFS open call
2327  * (nfs4_proc_file_open).
2328  *
2329  * The open for read will just hang around to be used by any process that
2330  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2331  */
2332
2333 static int
2334 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2335                  int flags, struct nameidata *nd)
2336 {
2337         struct path path = {
2338                 .mnt = nd->path.mnt,
2339                 .dentry = dentry,
2340         };
2341         struct nfs4_state *state;
2342         struct rpc_cred *cred;
2343         fmode_t fmode = flags & (FMODE_READ | FMODE_WRITE);
2344         int status = 0;
2345
2346         cred = rpc_lookup_cred();
2347         if (IS_ERR(cred)) {
2348                 status = PTR_ERR(cred);
2349                 goto out;
2350         }
2351         state = nfs4_do_open(dir, &path, fmode, flags, sattr, cred);
2352         d_drop(dentry);
2353         if (IS_ERR(state)) {
2354                 status = PTR_ERR(state);
2355                 goto out_putcred;
2356         }
2357         d_add(dentry, igrab(state->inode));
2358         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2359         if (flags & O_EXCL) {
2360                 struct nfs_fattr fattr;
2361                 status = nfs4_do_setattr(state->inode, cred, &fattr, sattr, state);
2362                 if (status == 0)
2363                         nfs_setattr_update_inode(state->inode, sattr);
2364                 nfs_post_op_update_inode(state->inode, &fattr);
2365         }
2366         if (status == 0 && (nd->flags & LOOKUP_OPEN) != 0)
2367                 status = nfs4_intent_set_file(nd, &path, state, fmode);
2368         else
2369                 nfs4_close_sync(&path, state, fmode);
2370 out_putcred:
2371         put_rpccred(cred);
2372 out:
2373         return status;
2374 }
2375
2376 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2377 {
2378         struct nfs_server *server = NFS_SERVER(dir);
2379         struct nfs_removeargs args = {
2380                 .fh = NFS_FH(dir),
2381                 .name.len = name->len,
2382                 .name.name = name->name,
2383                 .bitmask = server->attr_bitmask,
2384         };
2385         struct nfs_removeres res = {
2386                 .server = server,
2387         };
2388         struct rpc_message msg = {
2389                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2390                 .rpc_argp = &args,
2391                 .rpc_resp = &res,
2392         };
2393         int                     status;
2394
2395         nfs_fattr_init(&res.dir_attr);
2396         status = nfs4_call_sync(server, &msg, &args, &res, 1);
2397         if (status == 0) {
2398                 update_changeattr(dir, &res.cinfo);
2399                 nfs_post_op_update_inode(dir, &res.dir_attr);
2400         }
2401         return status;
2402 }
2403
2404 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2405 {
2406         struct nfs4_exception exception = { };
2407         int err;
2408         do {
2409                 err = nfs4_handle_exception(NFS_SERVER(dir),
2410                                 _nfs4_proc_remove(dir, name),
2411                                 &exception);
2412         } while (exception.retry);
2413         return err;
2414 }
2415
2416 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2417 {
2418         struct nfs_server *server = NFS_SERVER(dir);
2419         struct nfs_removeargs *args = msg->rpc_argp;
2420         struct nfs_removeres *res = msg->rpc_resp;
2421
2422         args->bitmask = server->cache_consistency_bitmask;
2423         res->server = server;
2424         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2425 }
2426
2427 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2428 {
2429         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2430
2431         nfs4_sequence_done(res->server, &res->seq_res, task->tk_status);
2432         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2433                 return 0;
2434         nfs4_sequence_free_slot(res->server->nfs_client, &res->seq_res);
2435         update_changeattr(dir, &res->cinfo);
2436         nfs_post_op_update_inode(dir, &res->dir_attr);
2437         return 1;
2438 }
2439
2440 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2441                 struct inode *new_dir, struct qstr *new_name)
2442 {
2443         struct nfs_server *server = NFS_SERVER(old_dir);
2444         struct nfs4_rename_arg arg = {
2445                 .old_dir = NFS_FH(old_dir),
2446                 .new_dir = NFS_FH(new_dir),
2447                 .old_name = old_name,
2448                 .new_name = new_name,
2449                 .bitmask = server->attr_bitmask,
2450         };
2451         struct nfs_fattr old_fattr, new_fattr;
2452         struct nfs4_rename_res res = {
2453                 .server = server,
2454                 .old_fattr = &old_fattr,
2455                 .new_fattr = &new_fattr,
2456         };
2457         struct rpc_message msg = {
2458                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2459                 .rpc_argp = &arg,
2460                 .rpc_resp = &res,
2461         };
2462         int                     status;
2463         
2464         nfs_fattr_init(res.old_fattr);
2465         nfs_fattr_init(res.new_fattr);
2466         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2467
2468         if (!status) {
2469                 update_changeattr(old_dir, &res.old_cinfo);
2470                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2471                 update_changeattr(new_dir, &res.new_cinfo);
2472                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2473         }
2474         return status;
2475 }
2476
2477 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2478                 struct inode *new_dir, struct qstr *new_name)
2479 {
2480         struct nfs4_exception exception = { };
2481         int err;
2482         do {
2483                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2484                                 _nfs4_proc_rename(old_dir, old_name,
2485                                         new_dir, new_name),
2486                                 &exception);
2487         } while (exception.retry);
2488         return err;
2489 }
2490
2491 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2492 {
2493         struct nfs_server *server = NFS_SERVER(inode);
2494         struct nfs4_link_arg arg = {
2495                 .fh     = NFS_FH(inode),
2496                 .dir_fh = NFS_FH(dir),
2497                 .name   = name,
2498                 .bitmask = server->attr_bitmask,
2499         };
2500         struct nfs_fattr fattr, dir_attr;
2501         struct nfs4_link_res res = {
2502                 .server = server,
2503                 .fattr = &fattr,
2504                 .dir_attr = &dir_attr,
2505         };
2506         struct rpc_message msg = {
2507                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2508                 .rpc_argp = &arg,
2509                 .rpc_resp = &res,
2510         };
2511         int                     status;
2512
2513         nfs_fattr_init(res.fattr);
2514         nfs_fattr_init(res.dir_attr);
2515         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2516         if (!status) {
2517                 update_changeattr(dir, &res.cinfo);
2518                 nfs_post_op_update_inode(dir, res.dir_attr);
2519                 nfs_post_op_update_inode(inode, res.fattr);
2520         }
2521
2522         return status;
2523 }
2524
2525 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2526 {
2527         struct nfs4_exception exception = { };
2528         int err;
2529         do {
2530                 err = nfs4_handle_exception(NFS_SERVER(inode),
2531                                 _nfs4_proc_link(inode, dir, name),
2532                                 &exception);
2533         } while (exception.retry);
2534         return err;
2535 }
2536
2537 struct nfs4_createdata {
2538         struct rpc_message msg;
2539         struct nfs4_create_arg arg;
2540         struct nfs4_create_res res;
2541         struct nfs_fh fh;
2542         struct nfs_fattr fattr;
2543         struct nfs_fattr dir_fattr;
2544 };
2545
2546 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2547                 struct qstr *name, struct iattr *sattr, u32 ftype)
2548 {
2549         struct nfs4_createdata *data;
2550
2551         data = kzalloc(sizeof(*data), GFP_KERNEL);
2552         if (data != NULL) {
2553                 struct nfs_server *server = NFS_SERVER(dir);
2554
2555                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2556                 data->msg.rpc_argp = &data->arg;
2557                 data->msg.rpc_resp = &data->res;
2558                 data->arg.dir_fh = NFS_FH(dir);
2559                 data->arg.server = server;
2560                 data->arg.name = name;
2561                 data->arg.attrs = sattr;
2562                 data->arg.ftype = ftype;
2563                 data->arg.bitmask = server->attr_bitmask;
2564                 data->res.server = server;
2565                 data->res.fh = &data->fh;
2566                 data->res.fattr = &data->fattr;
2567                 data->res.dir_fattr = &data->dir_fattr;
2568                 nfs_fattr_init(data->res.fattr);
2569                 nfs_fattr_init(data->res.dir_fattr);
2570         }
2571         return data;
2572 }
2573
2574 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2575 {
2576         int status = nfs4_call_sync(NFS_SERVER(dir), &data->msg,
2577                                     &data->arg, &data->res, 1);
2578         if (status == 0) {
2579                 update_changeattr(dir, &data->res.dir_cinfo);
2580                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2581                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2582         }
2583         return status;
2584 }
2585
2586 static void nfs4_free_createdata(struct nfs4_createdata *data)
2587 {
2588         kfree(data);
2589 }
2590
2591 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2592                 struct page *page, unsigned int len, struct iattr *sattr)
2593 {
2594         struct nfs4_createdata *data;
2595         int status = -ENAMETOOLONG;
2596
2597         if (len > NFS4_MAXPATHLEN)
2598                 goto out;
2599
2600         status = -ENOMEM;
2601         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2602         if (data == NULL)
2603                 goto out;
2604
2605         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2606         data->arg.u.symlink.pages = &page;
2607         data->arg.u.symlink.len = len;
2608         
2609         status = nfs4_do_create(dir, dentry, data);
2610
2611         nfs4_free_createdata(data);
2612 out:
2613         return status;
2614 }
2615
2616 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2617                 struct page *page, unsigned int len, struct iattr *sattr)
2618 {
2619         struct nfs4_exception exception = { };
2620         int err;
2621         do {
2622                 err = nfs4_handle_exception(NFS_SERVER(dir),
2623                                 _nfs4_proc_symlink(dir, dentry, page,
2624                                                         len, sattr),
2625                                 &exception);
2626         } while (exception.retry);
2627         return err;
2628 }
2629
2630 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2631                 struct iattr *sattr)
2632 {
2633         struct nfs4_createdata *data;
2634         int status = -ENOMEM;
2635
2636         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2637         if (data == NULL)
2638                 goto out;
2639
2640         status = nfs4_do_create(dir, dentry, data);
2641
2642         nfs4_free_createdata(data);
2643 out:
2644         return status;
2645 }
2646
2647 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2648                 struct iattr *sattr)
2649 {
2650         struct nfs4_exception exception = { };
2651         int err;
2652         do {
2653                 err = nfs4_handle_exception(NFS_SERVER(dir),
2654                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2655                                 &exception);
2656         } while (exception.retry);
2657         return err;
2658 }
2659
2660 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2661                   u64 cookie, struct page *page, unsigned int count, int plus)
2662 {
2663         struct inode            *dir = dentry->d_inode;
2664         struct nfs4_readdir_arg args = {
2665                 .fh = NFS_FH(dir),
2666                 .pages = &page,
2667                 .pgbase = 0,
2668                 .count = count,
2669                 .bitmask = NFS_SERVER(dentry->d_inode)->cache_consistency_bitmask,
2670         };
2671         struct nfs4_readdir_res res;
2672         struct rpc_message msg = {
2673                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2674                 .rpc_argp = &args,
2675                 .rpc_resp = &res,
2676                 .rpc_cred = cred,
2677         };
2678         int                     status;
2679
2680         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2681                         dentry->d_parent->d_name.name,
2682                         dentry->d_name.name,
2683                         (unsigned long long)cookie);
2684         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2685         res.pgbase = args.pgbase;
2686         status = nfs4_call_sync(NFS_SERVER(dir), &msg, &args, &res, 0);
2687         if (status == 0)
2688                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2689
2690         nfs_invalidate_atime(dir);
2691
2692         dprintk("%s: returns %d\n", __func__, status);
2693         return status;
2694 }
2695
2696 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2697                   u64 cookie, struct page *page, unsigned int count, int plus)
2698 {
2699         struct nfs4_exception exception = { };
2700         int err;
2701         do {
2702                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
2703                                 _nfs4_proc_readdir(dentry, cred, cookie,
2704                                         page, count, plus),
2705                                 &exception);
2706         } while (exception.retry);
2707         return err;
2708 }
2709
2710 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2711                 struct iattr *sattr, dev_t rdev)
2712 {
2713         struct nfs4_createdata *data;
2714         int mode = sattr->ia_mode;
2715         int status = -ENOMEM;
2716
2717         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
2718         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
2719
2720         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
2721         if (data == NULL)
2722                 goto out;
2723
2724         if (S_ISFIFO(mode))
2725                 data->arg.ftype = NF4FIFO;
2726         else if (S_ISBLK(mode)) {
2727                 data->arg.ftype = NF4BLK;
2728                 data->arg.u.device.specdata1 = MAJOR(rdev);
2729                 data->arg.u.device.specdata2 = MINOR(rdev);
2730         }
2731         else if (S_ISCHR(mode)) {
2732                 data->arg.ftype = NF4CHR;
2733                 data->arg.u.device.specdata1 = MAJOR(rdev);
2734                 data->arg.u.device.specdata2 = MINOR(rdev);
2735         }
2736         
2737         status = nfs4_do_create(dir, dentry, data);
2738
2739         nfs4_free_createdata(data);
2740 out:
2741         return status;
2742 }
2743
2744 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2745                 struct iattr *sattr, dev_t rdev)
2746 {
2747         struct nfs4_exception exception = { };
2748         int err;
2749         do {
2750                 err = nfs4_handle_exception(NFS_SERVER(dir),
2751                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
2752                                 &exception);
2753         } while (exception.retry);
2754         return err;
2755 }
2756
2757 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
2758                  struct nfs_fsstat *fsstat)
2759 {
2760         struct nfs4_statfs_arg args = {
2761                 .fh = fhandle,
2762                 .bitmask = server->attr_bitmask,
2763         };
2764         struct nfs4_statfs_res res = {
2765                 .fsstat = fsstat,
2766         };
2767         struct rpc_message msg = {
2768                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
2769                 .rpc_argp = &args,
2770                 .rpc_resp = &res,
2771         };
2772
2773         nfs_fattr_init(fsstat->fattr);
2774         return  nfs4_call_sync(server, &msg, &args, &res, 0);
2775 }
2776
2777 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
2778 {
2779         struct nfs4_exception exception = { };
2780         int err;
2781         do {
2782                 err = nfs4_handle_exception(server,
2783                                 _nfs4_proc_statfs(server, fhandle, fsstat),
2784                                 &exception);
2785         } while (exception.retry);
2786         return err;
2787 }
2788
2789 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
2790                 struct nfs_fsinfo *fsinfo)
2791 {
2792         struct nfs4_fsinfo_arg args = {
2793                 .fh = fhandle,
2794                 .bitmask = server->attr_bitmask,
2795         };
2796         struct nfs4_fsinfo_res res = {
2797                 .fsinfo = fsinfo,
2798         };
2799         struct rpc_message msg = {
2800                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
2801                 .rpc_argp = &args,
2802                 .rpc_resp = &res,
2803         };
2804
2805         return nfs4_call_sync(server, &msg, &args, &res, 0);
2806 }
2807
2808 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
2809 {
2810         struct nfs4_exception exception = { };
2811         int err;
2812
2813         do {
2814                 err = nfs4_handle_exception(server,
2815                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
2816                                 &exception);
2817         } while (exception.retry);
2818         return err;
2819 }
2820
2821 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
2822 {
2823         nfs_fattr_init(fsinfo->fattr);
2824         return nfs4_do_fsinfo(server, fhandle, fsinfo);
2825 }
2826
2827 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
2828                 struct nfs_pathconf *pathconf)
2829 {
2830         struct nfs4_pathconf_arg args = {
2831                 .fh = fhandle,
2832                 .bitmask = server->attr_bitmask,
2833         };
2834         struct nfs4_pathconf_res res = {
2835                 .pathconf = pathconf,
2836         };
2837         struct rpc_message msg = {
2838                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
2839                 .rpc_argp = &args,
2840                 .rpc_resp = &res,
2841         };
2842
2843         /* None of the pathconf attributes are mandatory to implement */
2844         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
2845                 memset(pathconf, 0, sizeof(*pathconf));
2846                 return 0;
2847         }
2848
2849         nfs_fattr_init(pathconf->fattr);
2850         return nfs4_call_sync(server, &msg, &args, &res, 0);
2851 }
2852
2853 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
2854                 struct nfs_pathconf *pathconf)
2855 {
2856         struct nfs4_exception exception = { };
2857         int err;
2858
2859         do {
2860                 err = nfs4_handle_exception(server,
2861                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
2862                                 &exception);
2863         } while (exception.retry);
2864         return err;
2865 }
2866
2867 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
2868 {
2869         struct nfs_server *server = NFS_SERVER(data->inode);
2870
2871         dprintk("--> %s\n", __func__);
2872
2873         /* nfs4_sequence_free_slot called in the read rpc_call_done */
2874         nfs4_sequence_done(server, &data->res.seq_res, task->tk_status);
2875
2876         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
2877                 rpc_restart_call(task);
2878                 return -EAGAIN;
2879         }
2880
2881         nfs_invalidate_atime(data->inode);
2882         if (task->tk_status > 0)
2883                 renew_lease(server, data->timestamp);
2884         return 0;
2885 }
2886
2887 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
2888 {
2889         data->timestamp   = jiffies;
2890         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
2891 }
2892
2893 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
2894 {
2895         struct inode *inode = data->inode;
2896         
2897         /* slot is freed in nfs_writeback_done */
2898         nfs4_sequence_done(NFS_SERVER(inode), &data->res.seq_res,
2899                            task->tk_status);
2900
2901         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
2902                 rpc_restart_call(task);
2903                 return -EAGAIN;
2904         }
2905         if (task->tk_status >= 0) {
2906                 renew_lease(NFS_SERVER(inode), data->timestamp);
2907                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
2908         }
2909         return 0;
2910 }
2911
2912 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
2913 {
2914         struct nfs_server *server = NFS_SERVER(data->inode);
2915
2916         data->args.bitmask = server->cache_consistency_bitmask;
2917         data->res.server = server;
2918         data->timestamp   = jiffies;
2919
2920         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
2921 }
2922
2923 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
2924 {
2925         struct inode *inode = data->inode;
2926         
2927         nfs4_sequence_done(NFS_SERVER(inode), &data->res.seq_res,
2928                            task->tk_status);
2929         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
2930                 rpc_restart_call(task);
2931                 return -EAGAIN;
2932         }
2933         nfs4_sequence_free_slot(NFS_SERVER(inode)->nfs_client,
2934                                 &data->res.seq_res);
2935         nfs_refresh_inode(inode, data->res.fattr);
2936         return 0;
2937 }
2938
2939 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
2940 {
2941         struct nfs_server *server = NFS_SERVER(data->inode);
2942         
2943         data->args.bitmask = server->cache_consistency_bitmask;
2944         data->res.server = server;
2945         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
2946 }
2947
2948 /*
2949  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
2950  * standalone procedure for queueing an asynchronous RENEW.
2951  */
2952 static void nfs4_renew_done(struct rpc_task *task, void *data)
2953 {
2954         struct nfs_client *clp = (struct nfs_client *)task->tk_msg.rpc_argp;
2955         unsigned long timestamp = (unsigned long)data;
2956
2957         if (task->tk_status < 0) {
2958                 /* Unless we're shutting down, schedule state recovery! */
2959                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
2960                         nfs4_schedule_state_recovery(clp);
2961                 return;
2962         }
2963         spin_lock(&clp->cl_lock);
2964         if (time_before(clp->cl_last_renewal,timestamp))
2965                 clp->cl_last_renewal = timestamp;
2966         spin_unlock(&clp->cl_lock);
2967 }
2968
2969 static const struct rpc_call_ops nfs4_renew_ops = {
2970         .rpc_call_done = nfs4_renew_done,
2971 };
2972
2973 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
2974 {
2975         struct rpc_message msg = {
2976                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
2977                 .rpc_argp       = clp,
2978                 .rpc_cred       = cred,
2979         };
2980
2981         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
2982                         &nfs4_renew_ops, (void *)jiffies);
2983 }
2984
2985 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
2986 {
2987         struct rpc_message msg = {
2988                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
2989                 .rpc_argp       = clp,
2990                 .rpc_cred       = cred,
2991         };
2992         unsigned long now = jiffies;
2993         int status;
2994
2995         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
2996         if (status < 0)
2997                 return status;
2998         spin_lock(&clp->cl_lock);
2999         if (time_before(clp->cl_last_renewal,now))
3000                 clp->cl_last_renewal = now;
3001         spin_unlock(&clp->cl_lock);
3002         return 0;
3003 }
3004
3005 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3006 {
3007         return (server->caps & NFS_CAP_ACLS)
3008                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3009                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3010 }
3011
3012 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3013  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3014  * the stack.
3015  */
3016 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3017
3018 static void buf_to_pages(const void *buf, size_t buflen,
3019                 struct page **pages, unsigned int *pgbase)
3020 {
3021         const void *p = buf;
3022
3023         *pgbase = offset_in_page(buf);
3024         p -= *pgbase;
3025         while (p < buf + buflen) {
3026                 *(pages++) = virt_to_page(p);
3027                 p += PAGE_CACHE_SIZE;
3028         }
3029 }
3030
3031 struct nfs4_cached_acl {
3032         int cached;
3033         size_t len;
3034         char data[0];
3035 };
3036
3037 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3038 {
3039         struct nfs_inode *nfsi = NFS_I(inode);
3040
3041         spin_lock(&inode->i_lock);
3042         kfree(nfsi->nfs4_acl);
3043         nfsi->nfs4_acl = acl;
3044         spin_unlock(&inode->i_lock);
3045 }
3046
3047 static void nfs4_zap_acl_attr(struct inode *inode)
3048 {
3049         nfs4_set_cached_acl(inode, NULL);
3050 }
3051
3052 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3053 {
3054         struct nfs_inode *nfsi = NFS_I(inode);
3055         struct nfs4_cached_acl *acl;
3056         int ret = -ENOENT;
3057
3058         spin_lock(&inode->i_lock);
3059         acl = nfsi->nfs4_acl;
3060         if (acl == NULL)
3061                 goto out;
3062         if (buf == NULL) /* user is just asking for length */
3063                 goto out_len;
3064         if (acl->cached == 0)
3065                 goto out;
3066         ret = -ERANGE; /* see getxattr(2) man page */
3067         if (acl->len > buflen)
3068                 goto out;
3069         memcpy(buf, acl->data, acl->len);
3070 out_len:
3071         ret = acl->len;
3072 out:
3073         spin_unlock(&inode->i_lock);
3074         return ret;
3075 }
3076
3077 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3078 {
3079         struct nfs4_cached_acl *acl;
3080
3081         if (buf && acl_len <= PAGE_SIZE) {
3082                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3083                 if (acl == NULL)
3084                         goto out;
3085                 acl->cached = 1;
3086                 memcpy(acl->data, buf, acl_len);
3087         } else {
3088                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3089                 if (acl == NULL)
3090                         goto out;
3091                 acl->cached = 0;
3092         }
3093         acl->len = acl_len;
3094 out:
3095         nfs4_set_cached_acl(inode, acl);
3096 }
3097
3098 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3099 {
3100         struct page *pages[NFS4ACL_MAXPAGES];
3101         struct nfs_getaclargs args = {
3102                 .fh = NFS_FH(inode),
3103                 .acl_pages = pages,
3104                 .acl_len = buflen,
3105         };
3106         struct nfs_getaclres res = {
3107                 .acl_len = buflen,
3108         };
3109         void *resp_buf;
3110         struct rpc_message msg = {
3111                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3112                 .rpc_argp = &args,
3113                 .rpc_resp = &res,
3114         };
3115         struct page *localpage = NULL;
3116         int ret;
3117
3118         if (buflen < PAGE_SIZE) {
3119                 /* As long as we're doing a round trip to the server anyway,
3120                  * let's be prepared for a page of acl data. */
3121                 localpage = alloc_page(GFP_KERNEL);
3122                 resp_buf = page_address(localpage);
3123                 if (localpage == NULL)
3124                         return -ENOMEM;
3125                 args.acl_pages[0] = localpage;
3126                 args.acl_pgbase = 0;
3127                 args.acl_len = PAGE_SIZE;
3128         } else {
3129                 resp_buf = buf;
3130                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3131         }
3132         ret = nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
3133         if (ret)
3134                 goto out_free;
3135         if (res.acl_len > args.acl_len)
3136                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3137         else
3138                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3139         if (buf) {
3140                 ret = -ERANGE;
3141                 if (res.acl_len > buflen)
3142                         goto out_free;
3143                 if (localpage)
3144                         memcpy(buf, resp_buf, res.acl_len);
3145         }
3146         ret = res.acl_len;
3147 out_free:
3148         if (localpage)
3149                 __free_page(localpage);
3150         return ret;
3151 }
3152
3153 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3154 {
3155         struct nfs4_exception exception = { };
3156         ssize_t ret;
3157         do {
3158                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3159                 if (ret >= 0)
3160                         break;
3161                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3162         } while (exception.retry);
3163         return ret;
3164 }
3165
3166 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3167 {
3168         struct nfs_server *server = NFS_SERVER(inode);
3169         int ret;
3170
3171         if (!nfs4_server_supports_acls(server))
3172                 return -EOPNOTSUPP;
3173         ret = nfs_revalidate_inode(server, inode);
3174         if (ret < 0)
3175                 return ret;
3176         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3177                 nfs_zap_acl_cache(inode);
3178         ret = nfs4_read_cached_acl(inode, buf, buflen);
3179         if (ret != -ENOENT)
3180                 return ret;
3181         return nfs4_get_acl_uncached(inode, buf, buflen);
3182 }
3183
3184 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3185 {
3186         struct nfs_server *server = NFS_SERVER(inode);
3187         struct page *pages[NFS4ACL_MAXPAGES];
3188         struct nfs_setaclargs arg = {
3189                 .fh             = NFS_FH(inode),
3190                 .acl_pages      = pages,
3191                 .acl_len        = buflen,
3192         };
3193         struct nfs_setaclres res;
3194         struct rpc_message msg = {
3195                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3196                 .rpc_argp       = &arg,
3197                 .rpc_resp       = &res,
3198         };
3199         int ret;
3200
3201         if (!nfs4_server_supports_acls(server))
3202                 return -EOPNOTSUPP;
3203         nfs_inode_return_delegation(inode);
3204         buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3205         ret = nfs4_call_sync(server, &msg, &arg, &res, 1);
3206         nfs_access_zap_cache(inode);
3207         nfs_zap_acl_cache(inode);
3208         return ret;
3209 }
3210
3211 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3212 {
3213         struct nfs4_exception exception = { };
3214         int err;
3215         do {
3216                 err = nfs4_handle_exception(NFS_SERVER(inode),
3217                                 __nfs4_proc_set_acl(inode, buf, buflen),
3218                                 &exception);
3219         } while (exception.retry);
3220         return err;
3221 }
3222
3223 static int
3224 _nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs_client *clp, struct nfs4_state *state)
3225 {
3226         if (!clp || task->tk_status >= 0)
3227                 return 0;
3228         switch(task->tk_status) {
3229                 case -NFS4ERR_ADMIN_REVOKED:
3230                 case -NFS4ERR_BAD_STATEID:
3231                 case -NFS4ERR_OPENMODE:
3232                         if (state == NULL)
3233                                 break;
3234                         nfs4_state_mark_reclaim_nograce(clp, state);
3235                 case -NFS4ERR_STALE_CLIENTID:
3236                 case -NFS4ERR_STALE_STATEID:
3237                 case -NFS4ERR_EXPIRED:
3238                         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3239                         nfs4_schedule_state_recovery(clp);
3240                         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3241                                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3242                         task->tk_status = 0;
3243                         return -EAGAIN;
3244                 case -NFS4ERR_DELAY:
3245                         if (server)
3246                                 nfs_inc_server_stats(server, NFSIOS_DELAY);
3247                 case -NFS4ERR_GRACE:
3248                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3249                         task->tk_status = 0;
3250                         return -EAGAIN;
3251                 case -NFS4ERR_OLD_STATEID:
3252                         task->tk_status = 0;
3253                         return -EAGAIN;
3254         }
3255         task->tk_status = nfs4_map_errors(task->tk_status);
3256         return 0;
3257 }
3258
3259 static int
3260 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3261 {
3262         return _nfs4_async_handle_error(task, server, server->nfs_client, state);
3263 }
3264
3265 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program, unsigned short port, struct rpc_cred *cred)
3266 {
3267         nfs4_verifier sc_verifier;
3268         struct nfs4_setclientid setclientid = {
3269                 .sc_verifier = &sc_verifier,
3270                 .sc_prog = program,
3271         };
3272         struct rpc_message msg = {
3273                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3274                 .rpc_argp = &setclientid,
3275                 .rpc_resp = clp,
3276                 .rpc_cred = cred,
3277         };
3278         __be32 *p;
3279         int loop = 0;
3280         int status;
3281
3282         p = (__be32*)sc_verifier.data;
3283         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3284         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3285
3286         for(;;) {
3287                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3288                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3289                                 clp->cl_ipaddr,
3290                                 rpc_peeraddr2str(clp->cl_rpcclient,
3291                                                         RPC_DISPLAY_ADDR),
3292                                 rpc_peeraddr2str(clp->cl_rpcclient,
3293                                                         RPC_DISPLAY_PROTO),
3294                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3295                                 clp->cl_id_uniquifier);
3296                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3297                                 sizeof(setclientid.sc_netid),
3298                                 rpc_peeraddr2str(clp->cl_rpcclient,
3299                                                         RPC_DISPLAY_NETID));
3300                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3301                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3302                                 clp->cl_ipaddr, port >> 8, port & 255);
3303
3304                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3305                 if (status != -NFS4ERR_CLID_INUSE)
3306                         break;
3307                 if (signalled())
3308                         break;
3309                 if (loop++ & 1)
3310                         ssleep(clp->cl_lease_time + 1);
3311                 else
3312                         if (++clp->cl_id_uniquifier == 0)
3313                                 break;
3314         }
3315         return status;
3316 }
3317
3318 static int _nfs4_proc_setclientid_confirm(struct nfs_client *clp, struct rpc_cred *cred)
3319 {
3320         struct nfs_fsinfo fsinfo;
3321         struct rpc_message msg = {
3322                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3323                 .rpc_argp = clp,
3324                 .rpc_resp = &fsinfo,
3325                 .rpc_cred = cred,
3326         };
3327         unsigned long now;
3328         int status;
3329
3330         now = jiffies;
3331         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3332         if (status == 0) {
3333                 spin_lock(&clp->cl_lock);
3334                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3335                 clp->cl_last_renewal = now;
3336                 spin_unlock(&clp->cl_lock);
3337         }
3338         return status;
3339 }
3340
3341 int nfs4_proc_setclientid_confirm(struct nfs_client *clp, struct rpc_cred *cred)
3342 {
3343         long timeout = 0;
3344         int err;
3345         do {
3346                 err = _nfs4_proc_setclientid_confirm(clp, cred);
3347                 switch (err) {
3348                         case 0:
3349                                 return err;
3350                         case -NFS4ERR_RESOURCE:
3351                                 /* The IBM lawyers misread another document! */
3352                         case -NFS4ERR_DELAY:
3353                                 err = nfs4_delay(clp->cl_rpcclient, &timeout);
3354                 }
3355         } while (err == 0);
3356         return err;
3357 }
3358
3359 struct nfs4_delegreturndata {
3360         struct nfs4_delegreturnargs args;
3361         struct nfs4_delegreturnres res;
3362         struct nfs_fh fh;
3363         nfs4_stateid stateid;
3364         unsigned long timestamp;
3365         struct nfs_fattr fattr;
3366         int rpc_status;
3367 };
3368
3369 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3370 {
3371         struct nfs4_delegreturndata *data = calldata;
3372
3373         nfs4_sequence_done_free_slot(data->res.server, &data->res.seq_res,
3374                                      task->tk_status);
3375
3376         data->rpc_status = task->tk_status;
3377         if (data->rpc_status == 0)
3378                 renew_lease(data->res.server, data->timestamp);
3379 }
3380
3381 static void nfs4_delegreturn_release(void *calldata)
3382 {
3383         kfree(calldata);
3384 }
3385
3386 #if defined(CONFIG_NFS_V4_1)
3387 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3388 {
3389         struct nfs4_delegreturndata *d_data;
3390
3391         d_data = (struct nfs4_delegreturndata *)data;
3392
3393         if (nfs4_setup_sequence(d_data->res.server->nfs_client,
3394                                 &d_data->args.seq_args,
3395                                 &d_data->res.seq_res, 1, task))
3396                 return;
3397         rpc_call_start(task);
3398 }
3399 #endif /* CONFIG_NFS_V4_1 */
3400
3401 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3402 #if defined(CONFIG_NFS_V4_1)
3403         .rpc_call_prepare = nfs4_delegreturn_prepare,
3404 #endif /* CONFIG_NFS_V4_1 */
3405         .rpc_call_done = nfs4_delegreturn_done,
3406         .rpc_release = nfs4_delegreturn_release,
3407 };
3408
3409 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3410 {
3411         struct nfs4_delegreturndata *data;
3412         struct nfs_server *server = NFS_SERVER(inode);
3413         struct rpc_task *task;
3414         struct rpc_message msg = {
3415                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3416                 .rpc_cred = cred,
3417         };
3418         struct rpc_task_setup task_setup_data = {
3419                 .rpc_client = server->client,
3420                 .rpc_message = &msg,
3421                 .callback_ops = &nfs4_delegreturn_ops,
3422                 .flags = RPC_TASK_ASYNC,
3423         };
3424         int status = 0;
3425
3426         data = kzalloc(sizeof(*data), GFP_KERNEL);
3427         if (data == NULL)
3428                 return -ENOMEM;
3429         data->args.fhandle = &data->fh;
3430         data->args.stateid = &data->stateid;
3431         data->args.bitmask = server->attr_bitmask;
3432         nfs_copy_fh(&data->fh, NFS_FH(inode));
3433         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3434         data->res.fattr = &data->fattr;
3435         data->res.server = server;
3436         data->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3437         nfs_fattr_init(data->res.fattr);
3438         data->timestamp = jiffies;
3439         data->rpc_status = 0;
3440
3441         task_setup_data.callback_data = data;
3442         msg.rpc_argp = &data->args,
3443         msg.rpc_resp = &data->res,
3444         task = rpc_run_task(&task_setup_data);
3445         if (IS_ERR(task))
3446                 return PTR_ERR(task);
3447         if (!issync)
3448                 goto out;
3449         status = nfs4_wait_for_completion_rpc_task(task);
3450         if (status != 0)
3451                 goto out;
3452         status = data->rpc_status;
3453         if (status != 0)
3454                 goto out;
3455         nfs_refresh_inode(inode, &data->fattr);
3456 out:
3457         rpc_put_task(task);
3458         return status;
3459 }
3460
3461 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3462 {
3463         struct nfs_server *server = NFS_SERVER(inode);
3464         struct nfs4_exception exception = { };
3465         int err;
3466         do {
3467                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3468                 switch (err) {
3469                         case -NFS4ERR_STALE_STATEID:
3470                         case -NFS4ERR_EXPIRED:
3471                         case 0:
3472                                 return 0;
3473                 }
3474                 err = nfs4_handle_exception(server, err, &exception);
3475         } while (exception.retry);
3476         return err;
3477 }
3478
3479 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3480 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3481
3482 /* 
3483  * sleep, with exponential backoff, and retry the LOCK operation. 
3484  */
3485 static unsigned long
3486 nfs4_set_lock_task_retry(unsigned long timeout)
3487 {
3488         schedule_timeout_killable(timeout);
3489         timeout <<= 1;
3490         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3491                 return NFS4_LOCK_MAXTIMEOUT;
3492         return timeout;
3493 }
3494
3495 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3496 {
3497         struct inode *inode = state->inode;
3498         struct nfs_server *server = NFS_SERVER(inode);
3499         struct nfs_client *clp = server->nfs_client;
3500         struct nfs_lockt_args arg = {
3501                 .fh = NFS_FH(inode),
3502                 .fl = request,
3503         };
3504         struct nfs_lockt_res res = {
3505                 .denied = request,
3506         };
3507         struct rpc_message msg = {
3508                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3509                 .rpc_argp       = &arg,
3510                 .rpc_resp       = &res,
3511                 .rpc_cred       = state->owner->so_cred,
3512         };
3513         struct nfs4_lock_state *lsp;
3514         int status;
3515
3516         arg.lock_owner.clientid = clp->cl_clientid;
3517         status = nfs4_set_lock_state(state, request);
3518         if (status != 0)
3519                 goto out;
3520         lsp = request->fl_u.nfs4_fl.owner;
3521         arg.lock_owner.id = lsp->ls_id.id;
3522         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
3523         switch (status) {
3524                 case 0:
3525                         request->fl_type = F_UNLCK;
3526                         break;
3527                 case -NFS4ERR_DENIED:
3528                         status = 0;
3529         }
3530         request->fl_ops->fl_release_private(request);
3531 out:
3532         return status;
3533 }
3534
3535 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3536 {
3537         struct nfs4_exception exception = { };
3538         int err;
3539
3540         do {
3541                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3542                                 _nfs4_proc_getlk(state, cmd, request),
3543                                 &exception);
3544         } while (exception.retry);
3545         return err;
3546 }
3547
3548 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3549 {
3550         int res = 0;
3551         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3552                 case FL_POSIX:
3553                         res = posix_lock_file_wait(file, fl);
3554                         break;
3555                 case FL_FLOCK:
3556                         res = flock_lock_file_wait(file, fl);
3557                         break;
3558                 default:
3559                         BUG();
3560         }
3561         return res;
3562 }
3563
3564 struct nfs4_unlockdata {
3565         struct nfs_locku_args arg;
3566         struct nfs_locku_res res;
3567         struct nfs4_lock_state *lsp;
3568         struct nfs_open_context *ctx;
3569         struct file_lock fl;
3570         const struct nfs_server *server;
3571         unsigned long timestamp;
3572 };
3573
3574 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
3575                 struct nfs_open_context *ctx,
3576                 struct nfs4_lock_state *lsp,
3577                 struct nfs_seqid *seqid)
3578 {
3579         struct nfs4_unlockdata *p;
3580         struct inode *inode = lsp->ls_state->inode;
3581
3582         p = kzalloc(sizeof(*p), GFP_KERNEL);
3583         if (p == NULL)
3584                 return NULL;
3585         p->arg.fh = NFS_FH(inode);
3586         p->arg.fl = &p->fl;
3587         p->arg.seqid = seqid;
3588         p->res.seqid = seqid;
3589         p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3590         p->arg.stateid = &lsp->ls_stateid;
3591         p->lsp = lsp;
3592         atomic_inc(&lsp->ls_count);
3593         /* Ensure we don't close file until we're done freeing locks! */
3594         p->ctx = get_nfs_open_context(ctx);
3595         memcpy(&p->fl, fl, sizeof(p->fl));
3596         p->server = NFS_SERVER(inode);
3597         return p;
3598 }
3599
3600 static void nfs4_locku_release_calldata(void *data)
3601 {
3602         struct nfs4_unlockdata *calldata = data;
3603         nfs_free_seqid(calldata->arg.seqid);
3604         nfs4_put_lock_state(calldata->lsp);
3605         put_nfs_open_context(calldata->ctx);
3606         kfree(calldata);
3607 }
3608
3609 static void nfs4_locku_done(struct rpc_task *task, void *data)
3610 {
3611         struct nfs4_unlockdata *calldata = data;
3612
3613         nfs4_sequence_done(calldata->server, &calldata->res.seq_res,
3614                            task->tk_status);
3615         if (RPC_ASSASSINATED(task))
3616                 return;
3617         switch (task->tk_status) {
3618                 case 0:
3619                         memcpy(calldata->lsp->ls_stateid.data,
3620                                         calldata->res.stateid.data,
3621                                         sizeof(calldata->lsp->ls_stateid.data));
3622                         renew_lease(calldata->server, calldata->timestamp);
3623                         break;
3624                 case -NFS4ERR_BAD_STATEID:
3625                 case -NFS4ERR_OLD_STATEID:
3626                 case -NFS4ERR_STALE_STATEID:
3627                 case -NFS4ERR_EXPIRED:
3628                         break;
3629                 default:
3630                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
3631                                 rpc_restart_call(task);
3632         }
3633         nfs4_sequence_free_slot(calldata->server->nfs_client,
3634                                 &calldata->res.seq_res);
3635 }
3636
3637 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
3638 {
3639         struct nfs4_unlockdata *calldata = data;
3640
3641         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3642                 return;
3643         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
3644                 /* Note: exit _without_ running nfs4_locku_done */
3645                 task->tk_action = NULL;
3646                 return;
3647         }
3648         calldata->timestamp = jiffies;
3649         if (nfs4_setup_sequence(calldata->server->nfs_client,
3650                                 &calldata->arg.seq_args,
3651                                 &calldata->res.seq_res, 1, task))
3652                 return;
3653         rpc_call_start(task);
3654 }
3655
3656 static const struct rpc_call_ops nfs4_locku_ops = {
3657         .rpc_call_prepare = nfs4_locku_prepare,
3658         .rpc_call_done = nfs4_locku_done,
3659         .rpc_release = nfs4_locku_release_calldata,
3660 };
3661
3662 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
3663                 struct nfs_open_context *ctx,
3664                 struct nfs4_lock_state *lsp,
3665                 struct nfs_seqid *seqid)
3666 {
3667         struct nfs4_unlockdata *data;
3668         struct rpc_message msg = {
3669                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
3670                 .rpc_cred = ctx->cred,
3671         };
3672         struct rpc_task_setup task_setup_data = {
3673                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
3674                 .rpc_message = &msg,
3675                 .callback_ops = &nfs4_locku_ops,
3676                 .workqueue = nfsiod_workqueue,
3677                 .flags = RPC_TASK_ASYNC,
3678         };
3679
3680         /* Ensure this is an unlock - when canceling a lock, the
3681          * canceled lock is passed in, and it won't be an unlock.
3682          */
3683         fl->fl_type = F_UNLCK;
3684
3685         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
3686         if (data == NULL) {
3687                 nfs_free_seqid(seqid);
3688                 return ERR_PTR(-ENOMEM);
3689         }
3690
3691         msg.rpc_argp = &data->arg,
3692         msg.rpc_resp = &data->res,
3693         task_setup_data.callback_data = data;
3694         return rpc_run_task(&task_setup_data);
3695 }
3696
3697 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
3698 {
3699         struct nfs_inode *nfsi = NFS_I(state->inode);
3700         struct nfs_seqid *seqid;
3701         struct nfs4_lock_state *lsp;
3702         struct rpc_task *task;
3703         int status = 0;
3704         unsigned char fl_flags = request->fl_flags;
3705
3706         status = nfs4_set_lock_state(state, request);
3707         /* Unlock _before_ we do the RPC call */
3708         request->fl_flags |= FL_EXISTS;
3709         down_read(&nfsi->rwsem);
3710         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
3711                 up_read(&nfsi->rwsem);
3712                 goto out;
3713         }
3714         up_read(&nfsi->rwsem);
3715         if (status != 0)
3716                 goto out;
3717         /* Is this a delegated lock? */
3718         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
3719                 goto out;
3720         lsp = request->fl_u.nfs4_fl.owner;
3721         seqid = nfs_alloc_seqid(&lsp->ls_seqid);
3722         status = -ENOMEM;
3723         if (seqid == NULL)
3724                 goto out;
3725         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
3726         status = PTR_ERR(task);
3727         if (IS_ERR(task))
3728                 goto out;
3729         status = nfs4_wait_for_completion_rpc_task(task);
3730         rpc_put_task(task);
3731 out:
3732         request->fl_flags = fl_flags;
3733         return status;
3734 }
3735
3736 struct nfs4_lockdata {
3737         struct nfs_lock_args arg;
3738         struct nfs_lock_res res;
3739         struct nfs4_lock_state *lsp;
3740         struct nfs_open_context *ctx;
3741         struct file_lock fl;
3742         unsigned long timestamp;
3743         int rpc_status;
3744         int cancelled;
3745         struct nfs_server *server;
3746 };
3747
3748 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
3749                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp)
3750 {
3751         struct nfs4_lockdata *p;
3752         struct inode *inode = lsp->ls_state->inode;
3753         struct nfs_server *server = NFS_SERVER(inode);
3754
3755         p = kzalloc(sizeof(*p), GFP_KERNEL);
3756         if (p == NULL)
3757                 return NULL;
3758
3759         p->arg.fh = NFS_FH(inode);
3760         p->arg.fl = &p->fl;
3761         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid);
3762         if (p->arg.open_seqid == NULL)
3763                 goto out_free;
3764         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid);
3765         if (p->arg.lock_seqid == NULL)
3766                 goto out_free_seqid;
3767         p->arg.lock_stateid = &lsp->ls_stateid;
3768         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
3769         p->arg.lock_owner.id = lsp->ls_id.id;
3770         p->res.lock_seqid = p->arg.lock_seqid;
3771         p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3772         p->lsp = lsp;
3773         p->server = server;
3774         atomic_inc(&lsp->ls_count);
3775         p->ctx = get_nfs_open_context(ctx);
3776         memcpy(&p->fl, fl, sizeof(p->fl));
3777         return p;
3778 out_free_seqid:
3779         nfs_free_seqid(p->arg.open_seqid);
3780 out_free:
3781         kfree(p);
3782         return NULL;
3783 }
3784
3785 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
3786 {
3787         struct nfs4_lockdata *data = calldata;
3788         struct nfs4_state *state = data->lsp->ls_state;
3789
3790         dprintk("%s: begin!\n", __func__);
3791         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
3792                 return;
3793         /* Do we need to do an open_to_lock_owner? */
3794         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
3795                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
3796                         return;
3797                 data->arg.open_stateid = &state->stateid;
3798                 data->arg.new_lock_owner = 1;
3799                 data->res.open_seqid = data->arg.open_seqid;
3800         } else
3801                 data->arg.new_lock_owner = 0;
3802         data->timestamp = jiffies;
3803         if (nfs4_setup_sequence(data->server->nfs_client, &data->arg.seq_args,
3804                                 &data->res.seq_res, 1, task))
3805                 return;
3806         rpc_call_start(task);
3807         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
3808 }
3809
3810 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
3811 {
3812         struct nfs4_lockdata *data = calldata;
3813
3814         dprintk("%s: begin!\n", __func__);
3815
3816         nfs4_sequence_done_free_slot(data->server, &data->res.seq_res,
3817                                      task->tk_status);
3818
3819         data->rpc_status = task->tk_status;
3820         if (RPC_ASSASSINATED(task))
3821                 goto out;
3822         if (data->arg.new_lock_owner != 0) {
3823                 if (data->rpc_status == 0)
3824                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
3825                 else
3826                         goto out;
3827         }
3828         if (data->rpc_status == 0) {
3829                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
3830                                         sizeof(data->lsp->ls_stateid.data));
3831                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
3832                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
3833         }
3834 out:
3835         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
3836 }
3837
3838 static void nfs4_lock_release(void *calldata)
3839 {
3840         struct nfs4_lockdata *data = calldata;
3841
3842         dprintk("%s: begin!\n", __func__);
3843         nfs_free_seqid(data->arg.open_seqid);
3844         if (data->cancelled != 0) {
3845                 struct rpc_task *task;
3846                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
3847                                 data->arg.lock_seqid);
3848                 if (!IS_ERR(task))
3849                         rpc_put_task(task);
3850                 dprintk("%s: cancelling lock!\n", __func__);
3851         } else
3852                 nfs_free_seqid(data->arg.lock_seqid);
3853         nfs4_put_lock_state(data->lsp);
3854         put_nfs_open_context(data->ctx);
3855         kfree(data);
3856         dprintk("%s: done!\n", __func__);
3857 }
3858
3859 static const struct rpc_call_ops nfs4_lock_ops = {
3860         .rpc_call_prepare = nfs4_lock_prepare,
3861         .rpc_call_done = nfs4_lock_done,
3862         .rpc_release = nfs4_lock_release,
3863 };
3864
3865 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int reclaim)
3866 {
3867         struct nfs4_lockdata *data;
3868         struct rpc_task *task;
3869         struct rpc_message msg = {
3870                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
3871                 .rpc_cred = state->owner->so_cred,
3872         };
3873         struct rpc_task_setup task_setup_data = {
3874                 .rpc_client = NFS_CLIENT(state->inode),
3875                 .rpc_message = &msg,
3876                 .callback_ops = &nfs4_lock_ops,
3877                 .workqueue = nfsiod_workqueue,
3878                 .flags = RPC_TASK_ASYNC,
3879         };
3880         int ret;
3881
3882         dprintk("%s: begin!\n", __func__);
3883         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
3884                         fl->fl_u.nfs4_fl.owner);
3885         if (data == NULL)
3886                 return -ENOMEM;
3887         if (IS_SETLKW(cmd))
3888                 data->arg.block = 1;
3889         if (reclaim != 0)
3890                 data->arg.reclaim = 1;
3891         msg.rpc_argp = &data->arg,
3892         msg.rpc_resp = &data->res,
3893         task_setup_data.callback_data = data;
3894         task = rpc_run_task(&task_setup_data);
3895         if (IS_ERR(task))
3896                 return PTR_ERR(task);
3897         ret = nfs4_wait_for_completion_rpc_task(task);
3898         if (ret == 0) {
3899                 ret = data->rpc_status;
3900                 if (ret == -NFS4ERR_DENIED)
3901                         ret = -EAGAIN;
3902         } else
3903                 data->cancelled = 1;
3904         rpc_put_task(task);
3905         dprintk("%s: done, ret = %d!\n", __func__, ret);
3906         return ret;
3907 }
3908
3909 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
3910 {
3911         struct nfs_server *server = NFS_SERVER(state->inode);
3912         struct nfs4_exception exception = { };
3913         int err;
3914
3915         do {
3916                 /* Cache the lock if possible... */
3917                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
3918                         return 0;
3919                 err = _nfs4_do_setlk(state, F_SETLK, request, 1);
3920                 if (err != -NFS4ERR_DELAY)
3921                         break;
3922                 nfs4_handle_exception(server, err, &exception);
3923         } while (exception.retry);
3924         return err;
3925 }
3926
3927 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
3928 {
3929         struct nfs_server *server = NFS_SERVER(state->inode);
3930         struct nfs4_exception exception = { };
3931         int err;
3932
3933         err = nfs4_set_lock_state(state, request);
3934         if (err != 0)
3935                 return err;
3936         do {
3937                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
3938                         return 0;
3939                 err = _nfs4_do_setlk(state, F_SETLK, request, 0);
3940                 if (err != -NFS4ERR_DELAY)
3941                         break;
3942                 nfs4_handle_exception(server, err, &exception);
3943         } while (exception.retry);
3944         return err;
3945 }
3946
3947 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3948 {
3949         struct nfs_inode *nfsi = NFS_I(state->inode);
3950         unsigned char fl_flags = request->fl_flags;
3951         int status;
3952
3953         /* Is this a delegated open? */
3954         status = nfs4_set_lock_state(state, request);
3955         if (status != 0)
3956                 goto out;
3957         request->fl_flags |= FL_ACCESS;
3958         status = do_vfs_lock(request->fl_file, request);
3959         if (status < 0)
3960                 goto out;
3961         down_read(&nfsi->rwsem);
3962         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
3963                 /* Yes: cache locks! */
3964                 /* ...but avoid races with delegation recall... */
3965                 request->fl_flags = fl_flags & ~FL_SLEEP;
3966                 status = do_vfs_lock(request->fl_file, request);
3967                 goto out_unlock;
3968         }
3969         status = _nfs4_do_setlk(state, cmd, request, 0);
3970         if (status != 0)
3971                 goto out_unlock;
3972         /* Note: we always want to sleep here! */
3973         request->fl_flags = fl_flags | FL_SLEEP;
3974         if (do_vfs_lock(request->fl_file, request) < 0)
3975                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
3976 out_unlock:
3977         up_read(&nfsi->rwsem);
3978 out:
3979         request->fl_flags = fl_flags;
3980         return status;
3981 }
3982
3983 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3984 {
3985         struct nfs4_exception exception = { };
3986         int err;
3987
3988         do {
3989                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3990                                 _nfs4_proc_setlk(state, cmd, request),
3991                                 &exception);
3992         } while (exception.retry);
3993         return err;
3994 }
3995
3996 static int
3997 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
3998 {
3999         struct nfs_open_context *ctx;
4000         struct nfs4_state *state;
4001         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4002         int status;
4003
4004         /* verify open state */
4005         ctx = nfs_file_open_context(filp);
4006         state = ctx->state;
4007
4008         if (request->fl_start < 0 || request->fl_end < 0)
4009                 return -EINVAL;
4010
4011         if (IS_GETLK(cmd))
4012                 return nfs4_proc_getlk(state, F_GETLK, request);
4013
4014         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4015                 return -EINVAL;
4016
4017         if (request->fl_type == F_UNLCK)
4018                 return nfs4_proc_unlck(state, cmd, request);
4019
4020         do {
4021                 status = nfs4_proc_setlk(state, cmd, request);
4022                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4023                         break;
4024                 timeout = nfs4_set_lock_task_retry(timeout);
4025                 status = -ERESTARTSYS;
4026                 if (signalled())
4027                         break;
4028         } while(status < 0);
4029         return status;
4030 }
4031
4032 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4033 {
4034         struct nfs_server *server = NFS_SERVER(state->inode);
4035         struct nfs4_exception exception = { };
4036         int err;
4037
4038         err = nfs4_set_lock_state(state, fl);
4039         if (err != 0)
4040                 goto out;
4041         do {
4042                 err = _nfs4_do_setlk(state, F_SETLK, fl, 0);
4043                 if (err != -NFS4ERR_DELAY)
4044                         break;
4045                 err = nfs4_handle_exception(server, err, &exception);
4046         } while (exception.retry);
4047 out:
4048         return err;
4049 }
4050
4051 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4052
4053 int nfs4_setxattr(struct dentry *dentry, const char *key, const void *buf,
4054                 size_t buflen, int flags)
4055 {
4056         struct inode *inode = dentry->d_inode;
4057
4058         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
4059                 return -EOPNOTSUPP;
4060
4061         return nfs4_proc_set_acl(inode, buf, buflen);
4062 }
4063
4064 /* The getxattr man page suggests returning -ENODATA for unknown attributes,
4065  * and that's what we'll do for e.g. user attributes that haven't been set.
4066  * But we'll follow ext2/ext3's lead by returning -EOPNOTSUPP for unsupported
4067  * attributes in kernel-managed attribute namespaces. */
4068 ssize_t nfs4_getxattr(struct dentry *dentry, const char *key, void *buf,
4069                 size_t buflen)
4070 {
4071         struct inode *inode = dentry->d_inode;
4072
4073         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
4074                 return -EOPNOTSUPP;
4075
4076         return nfs4_proc_get_acl(inode, buf, buflen);
4077 }
4078
4079 ssize_t nfs4_listxattr(struct dentry *dentry, char *buf, size_t buflen)
4080 {
4081         size_t len = strlen(XATTR_NAME_NFSV4_ACL) + 1;
4082
4083         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4084                 return 0;
4085         if (buf && buflen < len)
4086                 return -ERANGE;
4087         if (buf)
4088                 memcpy(buf, XATTR_NAME_NFSV4_ACL, len);
4089         return len;
4090 }
4091
4092 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4093 {
4094         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4095                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4096                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4097                 return;
4098
4099         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4100                 NFS_ATTR_FATTR_NLINK;
4101         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4102         fattr->nlink = 2;
4103 }
4104
4105 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4106                 struct nfs4_fs_locations *fs_locations, struct page *page)
4107 {
4108         struct nfs_server *server = NFS_SERVER(dir);
4109         u32 bitmask[2] = {
4110                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4111                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4112         };
4113         struct nfs4_fs_locations_arg args = {
4114                 .dir_fh = NFS_FH(dir),
4115                 .name = name,
4116                 .page = page,
4117                 .bitmask = bitmask,
4118         };
4119         struct nfs4_fs_locations_res res = {
4120                 .fs_locations = fs_locations,
4121         };
4122         struct rpc_message msg = {
4123                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4124                 .rpc_argp = &args,
4125                 .rpc_resp = &res,
4126         };
4127         int status;
4128
4129         dprintk("%s: start\n", __func__);
4130         nfs_fattr_init(&fs_locations->fattr);
4131         fs_locations->server = server;
4132         fs_locations->nlocations = 0;
4133         status = nfs4_call_sync(server, &msg, &args, &res, 0);
4134         nfs_fixup_referral_attributes(&fs_locations->fattr);
4135         dprintk("%s: returned status = %d\n", __func__, status);
4136         return status;
4137 }
4138
4139 #ifdef CONFIG_NFS_V4_1
4140 /*
4141  * nfs4_proc_exchange_id()
4142  *
4143  * Since the clientid has expired, all compounds using sessions
4144  * associated with the stale clientid will be returning
4145  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4146  * be in some phase of session reset.
4147  */
4148 static int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4149 {
4150         nfs4_verifier verifier;
4151         struct nfs41_exchange_id_args args = {
4152                 .client = clp,
4153                 .flags = clp->cl_exchange_flags,
4154         };
4155         struct nfs41_exchange_id_res res = {
4156                 .client = clp,
4157         };
4158         int status;
4159         struct rpc_message msg = {
4160                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4161                 .rpc_argp = &args,
4162                 .rpc_resp = &res,
4163                 .rpc_cred = cred,
4164         };
4165         __be32 *p;
4166
4167         dprintk("--> %s\n", __func__);
4168         BUG_ON(clp == NULL);
4169         p = (u32 *)verifier.data;
4170         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4171         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4172         args.verifier = &verifier;
4173
4174         while (1) {
4175                 args.id_len = scnprintf(args.id, sizeof(args.id),
4176                                         "%s/%s %u",
4177                                         clp->cl_ipaddr,
4178                                         rpc_peeraddr2str(clp->cl_rpcclient,
4179                                                          RPC_DISPLAY_ADDR),
4180                                         clp->cl_id_uniquifier);
4181
4182                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
4183
4184                 if (status != NFS4ERR_CLID_INUSE)
4185                         break;
4186
4187                 if (signalled())
4188                         break;
4189
4190                 if (++clp->cl_id_uniquifier == 0)
4191                         break;
4192         }
4193
4194         dprintk("<-- %s status= %d\n", __func__, status);
4195         return status;
4196 }
4197
4198 struct nfs4_get_lease_time_data {
4199         struct nfs4_get_lease_time_args *args;
4200         struct nfs4_get_lease_time_res *res;
4201         struct nfs_client *clp;
4202 };
4203
4204 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4205                                         void *calldata)
4206 {
4207         int ret;
4208         struct nfs4_get_lease_time_data *data =
4209                         (struct nfs4_get_lease_time_data *)calldata;
4210
4211         dprintk("--> %s\n", __func__);
4212         /* just setup sequence, do not trigger session recovery
4213            since we're invoked within one */
4214         ret = nfs41_setup_sequence(data->clp->cl_session,
4215                                         &data->args->la_seq_args,
4216                                         &data->res->lr_seq_res, 0, task);
4217
4218         BUG_ON(ret == -EAGAIN);
4219         rpc_call_start(task);
4220         dprintk("<-- %s\n", __func__);
4221 }
4222
4223 /*
4224  * Called from nfs4_state_manager thread for session setup, so don't recover
4225  * from sequence operation or clientid errors.
4226  */
4227 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4228 {
4229         struct nfs4_get_lease_time_data *data =
4230                         (struct nfs4_get_lease_time_data *)calldata;
4231
4232         dprintk("--> %s\n", __func__);
4233         nfs41_sequence_done(data->clp, &data->res->lr_seq_res, task->tk_status);
4234         switch (task->tk_status) {
4235         case -NFS4ERR_DELAY:
4236         case -NFS4ERR_GRACE:
4237                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4238                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4239                 task->tk_status = 0;
4240                 rpc_restart_call(task);
4241                 return;
4242         }
4243         nfs41_sequence_free_slot(data->clp, &data->res->lr_seq_res);
4244         dprintk("<-- %s\n", __func__);
4245 }
4246
4247 struct rpc_call_ops nfs4_get_lease_time_ops = {
4248         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4249         .rpc_call_done = nfs4_get_lease_time_done,
4250 };
4251
4252 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4253 {
4254         struct rpc_task *task;
4255         struct nfs4_get_lease_time_args args;
4256         struct nfs4_get_lease_time_res res = {
4257                 .lr_fsinfo = fsinfo,
4258         };
4259         struct nfs4_get_lease_time_data data = {
4260                 .args = &args,
4261                 .res = &res,
4262                 .clp = clp,
4263         };
4264         struct rpc_message msg = {
4265                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4266                 .rpc_argp = &args,
4267                 .rpc_resp = &res,
4268         };
4269         struct rpc_task_setup task_setup = {
4270                 .rpc_client = clp->cl_rpcclient,
4271                 .rpc_message = &msg,
4272                 .callback_ops = &nfs4_get_lease_time_ops,
4273                 .callback_data = &data
4274         };
4275         int status;
4276
4277         res.lr_seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
4278         dprintk("--> %s\n", __func__);
4279         task = rpc_run_task(&task_setup);
4280
4281         if (IS_ERR(task))
4282                 status = PTR_ERR(task);
4283         else {
4284                 status = task->tk_status;
4285                 rpc_put_task(task);
4286         }
4287         dprintk("<-- %s return %d\n", __func__, status);
4288
4289         return status;
4290 }
4291
4292 /* Reset a slot table */
4293 static int nfs4_reset_slot_table(struct nfs4_session *session)
4294 {
4295         struct nfs4_slot_table *tbl = &session->fc_slot_table;
4296         int i, max_slots = session->fc_attrs.max_reqs;
4297         int old_max_slots = session->fc_slot_table.max_slots;
4298         int ret = 0;
4299
4300         dprintk("--> %s: max_reqs=%u, tbl %p\n", __func__,
4301                 session->fc_attrs.max_reqs, tbl);
4302
4303         /* Until we have dynamic slot table adjustment, insist
4304          * upon the same slot table size */
4305         if (max_slots != old_max_slots) {
4306                 dprintk("%s reset slot table does't match old\n",
4307                         __func__);
4308                 ret = -EINVAL; /*XXX NFS4ERR_REQ_TOO_BIG ? */
4309                 goto out;
4310         }
4311         spin_lock(&tbl->slot_tbl_lock);
4312         for (i = 0; i < max_slots; ++i)
4313                 tbl->slots[i].seq_nr = 1;
4314         tbl->highest_used_slotid = -1;
4315         spin_unlock(&tbl->slot_tbl_lock);
4316         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4317                 tbl, tbl->slots, tbl->max_slots);
4318 out:
4319         dprintk("<-- %s: return %d\n", __func__, ret);
4320         return ret;
4321 }
4322
4323 /*
4324  * Initialize slot table
4325  */
4326 static int nfs4_init_slot_table(struct nfs4_session *session)
4327 {
4328         struct nfs4_slot_table *tbl = &session->fc_slot_table;
4329         int i, max_slots = session->fc_attrs.max_reqs;
4330         struct nfs4_slot *slot;
4331         int ret = -ENOMEM;
4332
4333         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
4334
4335         dprintk("--> %s: max_reqs=%u\n", __func__,
4336                 session->fc_attrs.max_reqs);
4337
4338         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_KERNEL);
4339         if (!slot)
4340                 goto out;
4341         for (i = 0; i < max_slots; ++i)
4342                 slot[i].seq_nr = 1;
4343         ret = 0;
4344
4345         spin_lock(&tbl->slot_tbl_lock);
4346         if (tbl->slots != NULL) {
4347                 spin_unlock(&tbl->slot_tbl_lock);
4348                 dprintk("%s: slot table already initialized. tbl=%p slots=%p\n",
4349                         __func__, tbl, tbl->slots);
4350                 WARN_ON(1);
4351                 goto out_free;
4352         }
4353         tbl->max_slots = max_slots;
4354         tbl->slots = slot;
4355         tbl->highest_used_slotid = -1;  /* no slot is currently used */
4356         spin_unlock(&tbl->slot_tbl_lock);
4357         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4358                 tbl, tbl->slots, tbl->max_slots);
4359 out:
4360         dprintk("<-- %s: return %d\n", __func__, ret);
4361         return ret;
4362 out_free:
4363         kfree(slot);
4364         goto out;
4365 }
4366
4367 /* Destroy the slot table */
4368 static void nfs4_destroy_slot_table(struct nfs4_session *session)
4369 {
4370         if (session->fc_slot_table.slots == NULL)
4371                 return;
4372         kfree(session->fc_slot_table.slots);
4373         session->fc_slot_table.slots = NULL;
4374         return;
4375 }
4376
4377 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
4378 {
4379         struct nfs4_session *session;
4380         struct nfs4_slot_table *tbl;
4381
4382         session = kzalloc(sizeof(struct nfs4_session), GFP_KERNEL);
4383         if (!session)
4384                 return NULL;
4385         tbl = &session->fc_slot_table;
4386         spin_lock_init(&tbl->slot_tbl_lock);
4387         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "Slot table");
4388         session->clp = clp;
4389         return session;
4390 }
4391
4392 void nfs4_destroy_session(struct nfs4_session *session)
4393 {
4394         nfs4_destroy_slot_table(session);
4395         kfree(session);
4396 }
4397
4398 /*
4399  * Initialize the values to be used by the client in CREATE_SESSION
4400  * If nfs4_init_session set the fore channel request and response sizes,
4401  * use them.
4402  *
4403  * Set the back channel max_resp_sz_cached to zero to force the client to
4404  * always set csa_cachethis to FALSE because the current implementation
4405  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
4406  */
4407 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
4408 {
4409         struct nfs4_session *session = args->client->cl_session;
4410         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
4411                      mxresp_sz = session->fc_attrs.max_resp_sz;
4412
4413         if (mxrqst_sz == 0)
4414                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
4415         if (mxresp_sz == 0)
4416                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
4417         /* Fore channel attributes */
4418         args->fc_attrs.headerpadsz = 0;
4419         args->fc_attrs.max_rqst_sz = mxrqst_sz;
4420         args->fc_attrs.max_resp_sz = mxresp_sz;
4421         args->fc_attrs.max_resp_sz_cached = mxresp_sz;
4422         args->fc_attrs.max_ops = NFS4_MAX_OPS;
4423         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
4424
4425         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
4426                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
4427                 __func__,
4428                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
4429                 args->fc_attrs.max_resp_sz_cached, args->fc_attrs.max_ops,
4430                 args->fc_attrs.max_reqs);
4431
4432         /* Back channel attributes */
4433         args->bc_attrs.headerpadsz = 0;
4434         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
4435         args->bc_attrs.max_resp_sz = PAGE_SIZE;
4436         args->bc_attrs.max_resp_sz_cached = 0;
4437         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
4438         args->bc_attrs.max_reqs = 1;
4439
4440         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
4441                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
4442                 __func__,
4443                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
4444                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
4445                 args->bc_attrs.max_reqs);
4446 }
4447
4448 static int _verify_channel_attr(char *chan, char *attr_name, u32 sent, u32 rcvd)
4449 {
4450         if (rcvd <= sent)
4451                 return 0;
4452         printk(KERN_WARNING "%s: Session INVALID: %s channel %s increased. "
4453                 "sent=%u rcvd=%u\n", __func__, chan, attr_name, sent, rcvd);
4454         return -EINVAL;
4455 }
4456
4457 #define _verify_fore_channel_attr(_name_) \
4458         _verify_channel_attr("fore", #_name_, \
4459                              args->fc_attrs._name_, \
4460                              session->fc_attrs._name_)
4461
4462 #define _verify_back_channel_attr(_name_) \
4463         _verify_channel_attr("back", #_name_, \
4464                              args->bc_attrs._name_, \
4465                              session->bc_attrs._name_)
4466
4467 /*
4468  * The server is not allowed to increase the fore channel header pad size,
4469  * maximum response size, or maximum number of operations.
4470  *
4471  * The back channel attributes are only negotiatied down: We send what the
4472  * (back channel) server insists upon.
4473  */
4474 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
4475                                      struct nfs4_session *session)
4476 {
4477         int ret = 0;
4478
4479         ret |= _verify_fore_channel_attr(headerpadsz);
4480         ret |= _verify_fore_channel_attr(max_resp_sz);
4481         ret |= _verify_fore_channel_attr(max_ops);
4482
4483         ret |= _verify_back_channel_attr(headerpadsz);
4484         ret |= _verify_back_channel_attr(max_rqst_sz);
4485         ret |= _verify_back_channel_attr(max_resp_sz);
4486         ret |= _verify_back_channel_attr(max_resp_sz_cached);
4487         ret |= _verify_back_channel_attr(max_ops);
4488         ret |= _verify_back_channel_attr(max_reqs);
4489
4490         return ret;
4491 }
4492
4493 static int _nfs4_proc_create_session(struct nfs_client *clp)
4494 {
4495         struct nfs4_session *session = clp->cl_session;
4496         struct nfs41_create_session_args args = {
4497                 .client = clp,
4498                 .cb_program = NFS4_CALLBACK,
4499         };
4500         struct nfs41_create_session_res res = {
4501                 .client = clp,
4502         };
4503         struct rpc_message msg = {
4504                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
4505                 .rpc_argp = &args,
4506                 .rpc_resp = &res,
4507         };
4508         int status;
4509
4510         nfs4_init_channel_attrs(&args);
4511         args.flags = (SESSION4_PERSIST);
4512
4513         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
4514
4515         if (!status)
4516                 /* Verify the session's negotiated channel_attrs values */
4517                 status = nfs4_verify_channel_attrs(&args, session);
4518         if (!status) {
4519                 /* Increment the clientid slot sequence id */
4520                 clp->cl_seqid++;
4521         }
4522
4523         return status;
4524 }
4525
4526 /*
4527  * Issues a CREATE_SESSION operation to the server.
4528  * It is the responsibility of the caller to verify the session is
4529  * expired before calling this routine.
4530  */
4531 int nfs4_proc_create_session(struct nfs_client *clp, int reset)
4532 {
4533         int status;
4534         unsigned *ptr;
4535         struct nfs_fsinfo fsinfo;
4536         struct nfs4_session *session = clp->cl_session;
4537
4538         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
4539
4540         status = _nfs4_proc_create_session(clp);
4541         if (status)
4542                 goto out;
4543
4544         /* Init or reset the fore channel */
4545         if (reset)
4546                 status = nfs4_reset_slot_table(session);
4547         else
4548                 status = nfs4_init_slot_table(session);
4549         dprintk("fore channel slot table initialization returned %d\n", status);
4550         if (status)
4551                 goto out;
4552
4553         ptr = (unsigned *)&session->sess_id.data[0];
4554         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
4555                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
4556
4557         if (reset)
4558                 /* Lease time is aleady set */
4559                 goto out;
4560
4561         /* Get the lease time */
4562         status = nfs4_proc_get_lease_time(clp, &fsinfo);
4563         if (status == 0) {
4564                 /* Update lease time and schedule renewal */
4565                 spin_lock(&clp->cl_lock);
4566                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4567                 clp->cl_last_renewal = jiffies;
4568                 clear_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
4569                 spin_unlock(&clp->cl_lock);
4570
4571                 nfs4_schedule_state_renewal(clp);
4572         }
4573 out:
4574         dprintk("<-- %s\n", __func__);
4575         return status;
4576 }
4577
4578 /*
4579  * Issue the over-the-wire RPC DESTROY_SESSION.
4580  * The caller must serialize access to this routine.
4581  */
4582 int nfs4_proc_destroy_session(struct nfs4_session *session)
4583 {
4584         int status = 0;
4585         struct rpc_message msg;
4586
4587         dprintk("--> nfs4_proc_destroy_session\n");
4588
4589         /* session is still being setup */
4590         if (session->clp->cl_cons_state != NFS_CS_READY)
4591                 return status;
4592
4593         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
4594         msg.rpc_argp = session;
4595         msg.rpc_resp = NULL;
4596         msg.rpc_cred = NULL;
4597         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
4598
4599         if (status)
4600                 printk(KERN_WARNING
4601                         "Got error %d from the server on DESTROY_SESSION. "
4602                         "Session has been destroyed regardless...\n", status);
4603
4604         dprintk("<-- nfs4_proc_destroy_session\n");
4605         return status;
4606 }
4607
4608 /*
4609  * Renew the cl_session lease.
4610  */
4611 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
4612 {
4613         struct nfs4_sequence_args args;
4614         struct nfs4_sequence_res res;
4615
4616         struct rpc_message msg = {
4617                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
4618                 .rpc_argp = &args,
4619                 .rpc_resp = &res,
4620                 .rpc_cred = cred,
4621         };
4622
4623         args.sa_cache_this = 0;
4624
4625         return nfs4_call_sync_sequence(clp, clp->cl_rpcclient, &msg, &args,
4626                                        &res, 0);
4627 }
4628
4629 void nfs41_sequence_call_done(struct rpc_task *task, void *data)
4630 {
4631         struct nfs_client *clp = (struct nfs_client *)data;
4632
4633         nfs41_sequence_done(clp, task->tk_msg.rpc_resp, task->tk_status);
4634
4635         if (task->tk_status < 0) {
4636                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
4637
4638                 if (_nfs4_async_handle_error(task, NULL, clp, NULL)
4639                                                                 == -EAGAIN) {
4640                         rpc_restart_call(task);
4641                         return;
4642                 }
4643         }
4644         nfs41_sequence_free_slot(clp, task->tk_msg.rpc_resp);
4645         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
4646
4647         put_rpccred(task->tk_msg.rpc_cred);
4648         kfree(task->tk_msg.rpc_argp);
4649         kfree(task->tk_msg.rpc_resp);
4650
4651         dprintk("<-- %s\n", __func__);
4652 }
4653
4654 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
4655 {
4656         struct nfs_client *clp;
4657         struct nfs4_sequence_args *args;
4658         struct nfs4_sequence_res *res;
4659
4660         clp = (struct nfs_client *)data;
4661         args = task->tk_msg.rpc_argp;
4662         res = task->tk_msg.rpc_resp;
4663
4664         if (nfs4_setup_sequence(clp, args, res, 0, task))
4665                 return;
4666         rpc_call_start(task);
4667 }
4668
4669 static const struct rpc_call_ops nfs41_sequence_ops = {
4670         .rpc_call_done = nfs41_sequence_call_done,
4671         .rpc_call_prepare = nfs41_sequence_prepare,
4672 };
4673
4674 static int nfs41_proc_async_sequence(struct nfs_client *clp,
4675                                      struct rpc_cred *cred)
4676 {
4677         struct nfs4_sequence_args *args;
4678         struct nfs4_sequence_res *res;
4679         struct rpc_message msg = {
4680                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
4681                 .rpc_cred = cred,
4682         };
4683
4684         args = kzalloc(sizeof(*args), GFP_KERNEL);
4685         if (!args)
4686                 return -ENOMEM;
4687         res = kzalloc(sizeof(*res), GFP_KERNEL);
4688         if (!res) {
4689                 kfree(args);
4690                 return -ENOMEM;
4691         }
4692         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
4693         msg.rpc_argp = args;
4694         msg.rpc_resp = res;
4695
4696         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
4697                               &nfs41_sequence_ops, (void *)clp);
4698 }
4699
4700 #endif /* CONFIG_NFS_V4_1 */
4701
4702 struct nfs4_state_recovery_ops nfs4_reboot_recovery_ops = {
4703         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
4704         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
4705         .recover_open   = nfs4_open_reclaim,
4706         .recover_lock   = nfs4_lock_reclaim,
4707 };
4708
4709 struct nfs4_state_recovery_ops nfs4_nograce_recovery_ops = {
4710         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
4711         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
4712         .recover_open   = nfs4_open_expired,
4713         .recover_lock   = nfs4_lock_expired,
4714 };
4715
4716 static const struct inode_operations nfs4_file_inode_operations = {
4717         .permission     = nfs_permission,
4718         .getattr        = nfs_getattr,
4719         .setattr        = nfs_setattr,
4720         .getxattr       = nfs4_getxattr,
4721         .setxattr       = nfs4_setxattr,
4722         .listxattr      = nfs4_listxattr,
4723 };
4724
4725 const struct nfs_rpc_ops nfs_v4_clientops = {
4726         .version        = 4,                    /* protocol version */
4727         .dentry_ops     = &nfs4_dentry_operations,
4728         .dir_inode_ops  = &nfs4_dir_inode_operations,
4729         .file_inode_ops = &nfs4_file_inode_operations,
4730         .getroot        = nfs4_proc_get_root,
4731         .getattr        = nfs4_proc_getattr,
4732         .setattr        = nfs4_proc_setattr,
4733         .lookupfh       = nfs4_proc_lookupfh,
4734         .lookup         = nfs4_proc_lookup,
4735         .access         = nfs4_proc_access,
4736         .readlink       = nfs4_proc_readlink,
4737         .create         = nfs4_proc_create,
4738         .remove         = nfs4_proc_remove,
4739         .unlink_setup   = nfs4_proc_unlink_setup,
4740         .unlink_done    = nfs4_proc_unlink_done,
4741         .rename         = nfs4_proc_rename,
4742         .link           = nfs4_proc_link,
4743         .symlink        = nfs4_proc_symlink,
4744         .mkdir          = nfs4_proc_mkdir,
4745         .rmdir          = nfs4_proc_remove,
4746         .readdir        = nfs4_proc_readdir,
4747         .mknod          = nfs4_proc_mknod,
4748         .statfs         = nfs4_proc_statfs,
4749         .fsinfo         = nfs4_proc_fsinfo,
4750         .pathconf       = nfs4_proc_pathconf,
4751         .set_capabilities = nfs4_server_capabilities,
4752         .decode_dirent  = nfs4_decode_dirent,
4753         .read_setup     = nfs4_proc_read_setup,
4754         .read_done      = nfs4_read_done,
4755         .write_setup    = nfs4_proc_write_setup,
4756         .write_done     = nfs4_write_done,
4757         .commit_setup   = nfs4_proc_commit_setup,
4758         .commit_done    = nfs4_commit_done,
4759         .lock           = nfs4_proc_lock,
4760         .clear_acl_cache = nfs4_zap_acl_attr,
4761         .close_context  = nfs4_close_context,
4762 };
4763
4764 /*
4765  * Local variables:
4766  *  c-basic-offset: 8
4767  * End:
4768  */